50D Suspicions Confirmed

[fike]

There are finally some test photos that are reasonably high-quality that confirm what many had suspected:  The 50D, with it's 50% more pixels has nearly identical high ISO performance as the 40D.

http://rolandlim.wordpress.com/2008/09/30/...eos-50d-review/

This is an improvement of a sort because they increased the MP without adversely affecting noise, but I think everyone hoped for enhanced image quality at high ISO.

So, now comes my next question:  

Q: Will noise reduction tools like neat image or noise ninja have better or worse results reducing noise from a higher resolution file.  

A; My inclination is to say there will be less loss of detail due to the normal smudging effect of noise reduction because there is more detail to start with in the higher resolution image.    

If my brokerage account doesn't take another 20% dive, I may just upgrade my 30D, but only if the price gets under $1200.

[JohnKoerner]

You sound like you're contradicting yourself.

If there is "less loss of detail" in the image ... then is that not the same as "enhanced image quality"  

Don't forget the enhanced LCD, weather sealing, and all that too ...

Jack



.

[fike]

You sound like you're contradicting yourself.

If there is "less loss of detail" in the image ... then is that not the same as "enhanced image quality"  

Don't forget the enhanced LCD, weather sealing, and all that too ...

Jack
.
[a href=\"index.php?act=findpost&pid=225849\"][{POST_SNAPBACK}][/a]

Have you looked at the comparisons at the link above?

It does seem rather contradictory doesn't it?  I think I am trying to apply some pretty subjective terms to a complex issue, but I'll try to clarify my intent.

If you use the same lens and have the same field of view between a 40D and a 50D, the 50D will have more pixiels, and the same ratio of noise to unit area of the image--I don't mean ratio of noise to pixels.  When I look at the comparison images, the grain of the noise or the blotchiness of the noise appears to be of the same magnitude, but that is when comparing the images at similar display sizes.

So that got me thinking that if you were to look at the 50D image at its full resolution, would the added image data allow you to do better noise reduction.  

I think this comes out most clearly when they compare 40D images to 50D images with the 40D uprezed to the 50D resolution.  In that case, the 50D has far more detail at the same ISO than the 40D.  

Will this extra detail allow me to apply noise reduction software without losing so much detail that its resolving power becomes merely equivalent or even worse than the 40D?

So, my hypothesis is that:

The 50D will clean-up better using noise reduction software than the equivalent 40D image.  

Now I need to get a camera and test that hypothesis.

[Ray]

It's a pity that Roland did not provide us with the full size images from the 40D so we could compare different parts of the image from say ISO 400 upwards.

What I'm seeing in these samples, is that ISO 3200 is not really usable on either camera, if one wants to maintain close to full image quality.

Of equal interest to me is the degree of image degradation at the high ISOs I use frequently, but use with some hesitation because I know there will be at least some loss of image quality with the 40D and 20D.

What I'm seeing here is that the loss of image quality with the 50D at ISO 3200, compared with ISO 1600, is far greater than the loss of image quality at ISO 1600, compared with ISO 400.

I think with the 50D I would have less hesitation in using ISO 1600. The sample 50D image at ISO 1600 looks virtually as clean and sharp as the ISO 400 shot. (Not quite at an extreme pixel-peeping level, but close enough.)

[BernardLanguillier]

There are finally some test photos that are reasonably high-quality that confirm what many had suspected:  The 50D, with it's 50% more pixels has nearly identical high ISO performance as the 40D.
[a href=\"index.php?act=findpost&pid=225815\"][{POST_SNAPBACK}][/a]

If that is true, it is already quite an achivement isn't it?

If low ISO image quality is higher, with roughly equivalent high ISO image quality, then it would seem to me that you have overall a better package.

Low and average ISO quality remains IMHO the highest priority for most people.

Cheers,
Bernard

[fike]

If that is true, it is already quite an achivement isn't it?

Cheers,
Bernard
[a href=\"index.php?act=findpost&pid=225905\"][{POST_SNAPBACK}][/a]

It certainly is.  Many folks have complained about this fact though.

[stever]

low-medium ISO image quality is excellent on every Nikon and Canon offering and most every other SLR

low noise at high ISO opens a large variety of image possibilities for wildlife, sports, candids, landscapes, flash illumination, ....

in addition it allows smaller and lighter smaller aperture lenses to be used instead of heavier, bulkier, and mor expensive optics

as we're already pushing on or surpassing the resolution capabilities of many (most?) SLR lenses, i'd rather have better high ISO performance than playing the megapixel game that is rampant in pocket cameras (of course i'd say the same for pocket cameras)

[mertmag]

Do this simple test and you will look at ACR and the 50D quite differently. Take a raw file shot at ISO 12800 and convert it with ACR (no adjustments) and then use the Canon utility to convert the file (just make sure you turn off the noise reduction in the Canon preferences). The Canon software will give you an astonishingly cleaner image than the Adobe ACR. I was shocked by the difference.

[The View]

If that is true, it is already quite an achivement isn't it?

If low ISO image quality is higher, with roughly equivalent high ISO image quality, then it would seem to me that you have overall a better package.

Low and average ISO quality remains IMHO the highest priority for most people.

Cheers,
Bernard

You are right!

I have the 40D, and am very happy with it.

But, obviosly, the 50D took another step.

It's strange, that whenever a new product comes out, you find people complaining about it. It's become a kind of entertainment: photo equipment bashing.

[The View]

Do this simple test and you will look at ACR and the 50D quite differently. Take a raw file shot at ISO 12800 and convert it with ACR (no adjustments) and then use the Canon utility to convert the file (just make sure you turn off the noise reduction in the Canon preferences). The Canon software will give you an astonishingly cleaner image than the Adobe ACR. I was shocked by the difference.

I also am a big fan of Canon's RAW converter, which is actually called Digital Photo Professional, short: DPP.

I am using it as my sole RAW converter at this time.

Especially in skin tones and describing light it is great.

The noise filter is extremely effective, but my 40D has such low noise levels at the ISO I shoot with, that I barely use it.

[Geoff Wittig]

Do this simple test and you will look at ACR and the 50D quite differently. Take a raw file shot at ISO 12800 and convert it with ACR (no adjustments) and then use the Canon utility to convert the file (just make sure you turn off the noise reduction in the Canon preferences). The Canon software will give you an astonishingly cleaner image than the Adobe ACR. I was shocked by the difference.

I could be wrong, but to my eye the difference between these two samples is simply imposition of very strong noise reduction by the Canon software. The ACR file leaves this up to you; it still has the usual grain-like chroma noise. The DPP version looks quite mushy with the typical artifacts of over-enthusiastic noise reduction. Even with noise reduction turned off, DPP is still obviously doing quite a bit of it behind the scenes.

[Ray]

Do this simple test and you will look at ACR and the 50D quite differently. Take a raw file shot at ISO 12800 and convert it with ACR (no adjustments) and then use the Canon utility to convert the file (just make sure you turn off the noise reduction in the Canon preferences). The Canon software will give you an astonishingly cleaner image than the Adobe ACR. I was shocked by the difference.

I rarely use DPP, but just out of curiosity I installed the DPP software that shipped with my 50D and compared an ISO 3200 shot, converting with ACR 4.6 and DPP 3.5 with all noise reduction and all sharpening turned off in both converters. (As a matter of fact, I can't see any sharpening options in DPP. Am I missing something?)

The results are very close. This shot was taken with the Canon 100-400 IS at F8, 400mm and 1/400th sec. It's not as sharp as it could be with a better lens. On the other hand, it's doubtful that the qualities of a better lens would be able to shine through the noise.

I don't know how you got such a vast difference at ISO 12800. There are two areas where noise reduction can be applied; default settings under preferences that apply to all images, and individual adjustments under "NR/Lens/ALO".

[attachment=8917:Full_images.jpg]  [attachment=8918:100__crops.jpg]

[Misirlou]

I rarely use DPP, but just out of curiosity I installed the DPP software that shipped with my 50D and compared an ISO 3200 shot, converting with ACR 4.6 and DPP 3.5 with all noise reduction and all sharpening turned off in both converters. (As a matter of fact, I can't see any sharpening options in DPP. Am I missing something?)

The results are very close. This shot was taken with the Canon 100-400 IS at F8, 400mm and 1/400th sec. It's not as sharp as it could be with a better lens. On the other hand, it's doubtful that the qualities of a better lens would be able to shine through the noise.

I don't know how you got such a vast difference at ISO 12800. There are two areas where noise reduction can be applied; default settings under preferences that apply to all images, and individual adjustments under "NR/Lens/ALO".

[attachment=8917:Full_images.jpg]  [attachment=8918:100__crops.jpg]

I'd really like to see that file run through DxO. Is there a DxO module out for the 50D and that lens yet?

[BruceHouston]

I'd really like to see that file run through DxO. Is there a DxO module out for the 50D and that lens yet?

I agree.  Very sharp image!

[situgrrl]

What I'm seeing in these samples, is that ISO 3200 is not really usable on either camera, if one wants to maintain close to full image quality.

Am I alone in not understanding this attitude?  Hi ISOs have always been there to "get the shot" rather than "maintain full image quality" - getting the shot - as supposed to failing to get the shot - IS full image quality.

7 years ago I was a rookie freelancer selling rock and roll photos to the UK press.  I used a EOS 1 with 28-70 and 80-200 f2.8 lenses, a spot meter and Fuji Press 800 pushed to anywhere from 1600 to the end of the earth.  The film went to Ceta (memories!) for 1 hr processing, was scanned and emailed to the various picture desks.  On a good day, my 35mm neg was blown up for a single page spread - even occasionally after all hell being cropped out of the image.  No one ever called to say "could you shoot a lower ISO - the grain is awful" - the limitations of the technology were recognized, the file pushed as hard as possible and the best result possible was accepted.  I would always shoot the lowest ISO possible - given that I could not adjust it shot-shot like is now possible - and when the light brightened by 4 stops just as I'd rattled off the first 12 frames at 6400 - I swore, stopped the lens down and carried on.

With digital cameras reaching 8 MP we were able to get A3 enlargements at a quality previously considered to be medium format territory.  At the same time, we were given high ISO capabilities that allowed us to shoot at 1600 with similar quality to ISO 400 film - which was considered "every day" speed by 2000.  Whereas heavily pushed film would reduce the DR and up contrast so that there was only 3-5 stops, with digital, you loose only a stop or 2 off a medium with much greater range anyhow.  More than this, with PS plugins like Noiseware, noise is now something that can be "fixed in post" rather than a more or less inherent aspect of final image quality.

7-8 years ago it was possible with the right film, an excellent film processor and a deal of luck to take a grainy picture which hopefully conveyed enough atmosphere to be worthy of stretching to an A4 print.  From the specs of a 50D (and I've not used one) it sounds to me like you have the ability to produce an image far wider in dynamic range at any given ISO (especially over 800) with a native A3 ish file size - ie - with a 50D you can produce a poster sized image with minimal post-processing from a high ISO image that will be technically and aesthetically far superior to even a 5x7 print off film - and which was quite acceptable for national publication less than a decade ago.....

And you want more!  What do you need it for?!

[Panopeeper]

From the specs of a 50D (and I've not used one) it sounds to me like you have the ability to produce an image far wider in dynamic range at any given ISO (especially over 800)

Wider than what? Not wider than with the 40D.

And you want more!  What do you need it for?!

I got used to this funny "argument" from paper dinosaurs, but not from someone, who has been a rookie for seven years ago.

When I was born there was not only no internet but not even TV. I wonder why you are using the newest technology to post your arguments instead of mailing them to the members in letters.

Anyway, to your question, why some want more: I want higher DR, and lower noise is one of the ways to achieve that. Instead, Canon decided for more pixels what I have no use for.

[situgrrl]

Wider than what? Not wider than with the 40D.

No - I was arguing from the point of film - something I accept it moot for most people here - but also something I still use more than digital simply because of the "form factor" (I don't get along with SLRs generally)

I got used to this funny "argument" from paper dinosaurs, but not from someone, who has been a rookie for seven years ago.

When I was born there was not only no internet but not even TV. I wonder why you are using the newest technology to post your arguments instead of mailing them to the members in letters.

Thank you for suggesting I'm a philistine!  I will clarify my opinion - and that is all it is.  

It is my opinion that high ISO performance has got to a stage far beyond "good enough" which, for people like me who are essentially nocturnal is fantastic news.  Please don't confuse this with wanting development to stop!  I can see genuine uses for ISO 25000 and very much look forward to owning a camera with such an option.  I suspect that at this sensitivity, there remain some more serious noise issues than lie at "normal" high ISO values (800-3200) and I hope in future itinerations that it is reduced to a more preferable level.  That said, if I can print at 11x14 at this speed, I'll be stoked and if I can go bigger....cool!

"Preferable level" is I guess what is up for debate here - and as I see it, in the 800-3200 realm there is little to complain about - it is possible to capture an image which can be printed at A3 with "35mm type" equipment.  That's not to say I don't want less noise - it's just that I think there are more important things for manufacturers to concentrate at this point.  Ergonomics, size and weight for instance.  (All 3 explain why I'm back to film for my main camera - and even then, I'd swap my M4-P for an M8 the moment a winning lottery ticket comes my way).  

In likelihood, my next digital camera will be micro 4/3rds - perhaps ironic given 4/3rds problems with noise up until now - however - like I said, a DSLR is a heavy beast and I used to leave mine at home more than I took it.....and it's "getting the shot" that counts above all else.  Were I stronger, if camera bags matched my outfit better, things might be different!

Anyway, to your question, why some want more: I want higher DR, and lower noise is one of the ways to achieve that. Instead, Canon decided for more pixels what I have no use for.

Wanting more DR seems very reasonable - as someone who shoots perhaps 85% in B&W it was perhaps easier for me than a colour shooter to return to film.  The simple fact is that any digital file I have processed to black and white from RAW lacks the tonality of a well processed negative.  As a result, when I was shooting digital, I found myself stripping around 90% of the saturation from the picture rather than converting to true B&W fairly frequently.

I understand that a high pixel density leads to high noise without underlying geeky stuff coming into play.  I understand that many expected the 15 MP 50D to perform worse in this respect to the 40D - and I understand that many are therefore pleasantly surprised that it is the same.  I don't understand though why a higher pixel density affects the dynamic range.  Surely that is determined by the range of sensitivity at any given gain setting?  It seems logical that at higher ISOs, noise limits DR by blocking up shadows in particular but this is not an issue at lower ISOs - surely where the extended DR is noticeable?  I don't understand how reducing a chip's noise level at 3200 would translate to more DR at 100 - where cameras have been essentially noise free for sometime.

[Panopeeper]

Thank you for suggesting I'm a philistine!

I would not have come to say that. It is rather "inconsisteny" in my eyes: using the newest technology while condemning those, who want more development.

It is my opinion that high ISO performance has got to a stage far beyond "good enough" which, for people like me who are essentially nocturnal is fantastic news

Here is the bad news: it is far from being good enough for some other people (that's not me). The Nikon D3 (and D700) is about 1 2/3 stops better than the 40D (which is ~ 50D), and that is still not enough. I saw shots for example from jazz concert, really low light, so noisy, that the only way to accept it is saying "that's the best the camera can do".

Furthermore, I am pretty sure that you would find cases in your own experience, when you would have made the shot with a faster shutter in order to eliminate motion blur.

I can see genuine uses for ISO 25000 and very much look forward to owning a camera with such an option

Even worse news: there is no such camera (commercionally available). The Nikon D3 (the best at the moment) goes up to 6400 only. 12800 and 25600 is simply underexposure. The 50D too goes only up to 3200.

The simple fact is that any digital file I have processed to black and white from RAW lacks the tonality of a well processed negative.  As a result, when I was shooting digital, I found myself stripping around 90% of the saturation from the picture rather than converting to true B&W fairly frequently

I am not convinced, that this is a handicap of digital cameras per se. I tend to believe, that it is caused by the processing, i.e. by the b&w conversion. Still, I can imagine future b&w dedicated cameras, i.e. such without color filters over the sensels. (A side effect: this would increase the sensitivity perhaps by 100%.)

I don't understand though why a higher pixel density affects the dynamic range.  Surely that is determined by the range of sensitivity at any given gain setting?  It seems logical that at higher ISOs, noise limits DR by blocking up shadows in particular but this is not an issue at lower ISOs - surely where the extended DR is noticeable?  I don't understand how reducing a chip's noise level at 3200 would translate to more DR at 100 - where cameras have been essentially noise free for sometime.

It is correct, that the DR is the highest at the base (native) ISO, 100 or 200 or so. However, if the noise is low at high ISO, then presumably it is low at low ISO as well.

An important difference between film and digital sensor is, that the latter records linear data and that the recording capacity ends abruptly. Thus the upper end of the dynamic range is always clear cut. (When I am measuring/comparing the DR, I do not care for the highlights, I need only very dark spots.) However, the low end of the DR is mushy. It is not limited by zero pixel value, but by the noise, which exceeds acceptability far over zero. The lower the noise, the deeper one can go into the shadows (closer to the pixel value zero).

I guess you too encountered the stituation (with digital), when the sky was already burning and some dark areas were still close to total black.

[Ray]

Am I alone in not understanding this attitude?  Hi ISOs have always been there to "get the shot" rather than "maintain full image quality" - getting the shot - as supposed to failing to get the shot - IS full image quality.

I think you have not understood the situation. Merely getting the shot is more a concern of snapshooters and press photographers (although I hesitate to put them in the same category).

I'm not on assignment when I take my camera with me. I'm free to shoot whatever I find interesting, and having decided that some scene, activity, person etc could be the subject of an interesting photo, I want that photo to be technically as high quality as the circumstances and light conditions allow.

A high ISO setting is not just about getting an exposure in poor lighting conditions which looks okay, it's also about being able to use a sufficiently fast shutter speed to get a sharp image. When such attempted sharpness is obscured by noise, as it used to be with high-ISO film, and always is obscured with current P&S digital cameras at ISO 400 and above, then I'm not happy

[ErikKaffehr]

I certainly feel that both high ISO and IS are most helpful. I prefer to put my camera on tripod, whenever practical, but carrying a tripod is a nuisance at best (I normally have a CF tripod strapped to my backpack). A tripod can eliminate or reduce camera movement but doesn't help with a moving target and the same applies for IS.

I generally found that color slide film was not usable above 100 ISO, so shooting digital at 200-800 ISO is an improvement to me.

Regarding the really high ISO-s it seems that there is a lot of "scientific evidence" that real ISO flattens out around 1000 or so (based on "unity gain" and other analysis) but there are obviously many experienced shooters on these forums who can use high (and probably fake) ISO-s and get good results. It's a little bit like "pushing" film in the old days, never worked in theory but pretty well in practice ;-)

One additional thought is that using high ISO with ETTR (Expose To The Right) is like using lower ISO without ETTR.

Erik

I think you have not understood the situation. Merely getting the shot is more a concern of snapshooters and press photographers (although I hesitate to put them in the same category).

I'm not on assignment when I take my camera with me. I'm free to shoot whatever I find interesting, and having decided that some scene, activity, person etc could be the subject of an interesting photo, I want that photo to be technically as high quality as the circumstances and light conditions allow.

A high ISO setting is not just about getting an exposure in poor lighting conditions which looks okay, it's also about being able to use a sufficiently fast shutter speed to get a sharp image. When such attempted sharpness is obscured by noise, as it used to be with high-ISO film, and always is obscured with current P&S digital cameras at ISO 400 and above, then I'm not happy