hi-speed ASA noise differences for Canon cameras

[Gregory]

when photographing birds in my neighbourhood, I usually find myself fighting to get the necessary shutter speeds to produce crisper photos. here in Hong Kong, we're under the edge of the Asian Brown Cloud (not sure if it's fact or fiction, but I personally think it's fact) and even during the day, light can be relatively low, particularly under the shade of trees where many of the birds remain during the heat of the day, coming out only in the early morning and late evening before sunset.

even at ASA400, my shutter speeds can be dismally low. today, I was shooting with just 1/30th of a second, far below what is necessary to shoot these wonderful creatures that can effortlessly turn and jump at a seemingly lightning speed.

(background: for many of my photos, I'm shooting with either a 70-200/2.8 non-IS + 2x or a 70-300 DO IS, with an aperture of 6.3 which is just ok to get enough of the bird in focus at the typical distances of 10 to 15 feet -- DOF approx 2" according to the online DOF calculator)

ASA800 would be helpful, even ASA1600, but I feel that 800 is too noisy on my 350D. maybe others have different reactions and it would be great to hear from them.

I'd like to know if anyone has compared the 800 and 1600 noise of the 350D, 5D and 1Ds cameras. unfortunately with the 5D and 1Ds cameras, I lose the 1.6x multiplier, therefore requiring longer lenses, but it might be a necessary sacrifice in dimmer environments.

(btw, thanks to everyone who has answered or contributed to my recent topics. the answers have been very helpful and educational. they're much appreciated.)

regards,
Gregory

[jani]

I'd like to know if anyone has compared the 800 and 1600 noise of the 350D, 5D and 1Ds cameras. unfortunately with the 5D and 1Ds cameras, I lose the 1.6x multiplier, therefore requiring longer lenses, but it might be a necessary sacrifice in dimmer environments.

I'm not aware of any direct comparisons on the web, but if you look at the reviews on DPReview.com, you'll be able to compare for yourself:

EOS 350/Digital Rebel XT
EOS 20D
EOS 5D
EOS 1Ds
EOS 1Ds MkII

[Ray]

I'd like to know if anyone has compared the 800 and 1600 noise of the 350D, 5D and 1Ds cameras. unfortunately with the 5D and 1Ds cameras, I lose the 1.6x multiplier, therefore requiring longer lenses, but it might be a necessary sacrifice in dimmer environments.
[a href=\"index.php?act=findpost&pid=71636\"][{POST_SNAPBACK}][/a]

I have a 20D, which would have similar specs to the 350D, and a 5D. I don't find any significant difference in noise levels between the two cameras, although I suspect the 5D would produce less noise on a print of equal size as a result of its greater number of pixels.

When comparing the Dpreview noise tests, you should bear in mind that the comparisons are of equal size crops consisting of an equal number of pixels. It's essentially the noise of each pixel that's being compared. But as Jack Flesher would probably point out, a real world 5D image is likely to consist of around 12mp and a real world 350D image, 8mp. Whetever size print we make, the 350D image will be interpolated to a greater degree than the 5D image, together with its inherent noise, or downsampled to a smaller extent than the 5D image.

I find that ISO 1600 is perfectly usable on both cameras, provided you expose fully to the right. If you are shooting a bird in the shade of a tree, with a telephoto lens, then ISO 1600 is the setting to use. If there are patches of sky appearing through the foliage, then of course those would have to be blown highlights. Expose for the bird.

[fnagy]

Gregory,

I have had a chance to use the 20D and the 5D as well as the 1DsII (which is the camera of choice for me at this time), I shoot a lot of musicians at night and dimm clubs, files from the 1DSII are consistently cleaner at 800 and 1600, the 5D a is close but no cigar IMHO, and the auto fucus is slower in those situations.

Rent and shoot all three and make your selection based on need and budget?

[John Sheehy]

Gregory,

I have had a chance to use the 20D and the 5D as well as the 1DsII (which is the camera of choice for me at this time), I shoot a lot of musicians at night and dimm clubs, files from the 1DSII are consistently cleaner at 800 and 1600, the 5D a is close but no cigar IMHO, and the auto fucus is slower in those situations.

I'm surprised by that statement.  I've never heard anyone say that before.  The 5D is generally considered the least noisey, and the 1DsmkII slightly noisier than the 20D at ISO 1600, if my memory of previous comments is correct.

[added in edit] Perhaps I am thinking of "at 100%" comments; that could be the difference.

Rent and shoot all three and make your selection based on need and budget?
[a href=\"index.php?act=findpost&pid=71917\"][{POST_SNAPBACK}][/a]

Does anyone know if there is a collection anywhere of RAW blackframes and out-of-focus step wedges or color checkers for various cameras and ISOs?  RAW doesn't lie - except that the actual saturation exposure can vary from camera to camera. but at least you can see what the DR is like.

Of course, the only fair way to compare is to equalize exposure index by using the same lens, f-stop, and controlled exposure time, such as manual flash.  If one camera has an actual EI of 1200 when set to 1600, and another 2000, the 1200 one is going to have an unfair noise advantage.

[gochugogi]

I have a 5D and 20D and, although both are excellent at high ISO, I find the 5D images far superior in this aspect. Why? To make a 8 x 12 or 12 x 18 print the 20D file must be enlarged more than 5D thereby revealing more defects/artifacts. The 5D images simply look smoother and more detailed at the same print sizes. I suspect pixel to pixel both bodies are similar in noise but a FF sensor doesn't need to be enlarged nearly as much as a cropped one.

[digitaldog]

Often I use 1600 on my Rebel (RAW) under low llight because it looks so awesome to capture a scene with available light. I'm just shocked at how beautiful the lighting is and amazed at what I can capture that I'd never get with film. Noise, sure. But it's not that bad and you should really check out Noiseware plug in which is absolutely astoundingly at removing noise while keeping detail. Shocking and worth every penny. I output test images I shot at 1600 with and without the Noiseware reduction to show my wife (not a photo expert). I asked her to tell me which print she liked and why. Shy spotted the noise in a second and preferred the filtered prints.

Process the RAWs as you wish but turn off noise reduction and sharpening in the converter (I used Lightroom). Then run Noiseware filter and lastly capture sharpen. The order here is important.

I'm really looking forward to getting a 5D and going to 3200 <g>.

[John Sheehy]

I'm really looking forward to getting a 5D and going to 3200 <g>.
[a href=\"index.php?act=findpost&pid=71990\"][{POST_SNAPBACK}][/a]

Is the 5D's 3200 real?  I doubt it, as there are no ISOs between 1600 and 3200, and 3200 is an "ISO Expansion".

3200 will finally come of age when it is purely gain-based.  Right now, it's just pushed 1600 in most, if not all DSLRs, and exactly twice the noise of 1600.  A real 3200 would probably only have about 1.6x as much noise as 1600.

[jani]

Is the 5D's 3200 real?  I doubt it, as there are no ISOs between 1600 and 3200, and 3200 is an "ISO Expansion".

That's right, it's not "real". But it means that if you want to underexpose with three stops, you can do it (the limit for exposure compensation is +/- 2 Ev on any non-1-series digital; the 1-series have +/- 3 Ev).

[John Sheehy]

That's right, it's not "real". But it means that if you want to underexpose with three stops, you can do it (the limit for exposure compensation is +/- 2 Ev on any non-1-series digital; the 1-series have +/- 3 Ev).
[a href=\"index.php?act=findpost&pid=72051\"][{POST_SNAPBACK}][/a]

That's a theoretical benefit, which may or may not be of help in actual photography.  I rarely use negative exposure compensation, and when I do, it is usually no stronger than -1.  Perhaps a sunlit egret against a shaded background might warrant a -1.3, but I will often switch to manual and use a "sunny f/14" rule for that.

Where I want more EC is in the positive range.  +2 is extremely feeble, to say the least.  A white wall at +2 will result in a RAW file with green as the strongest channel, with about 1400 RAW levels as a peak in the green channel out of 3967 with my 20D; still 1.5 stops from clipping the green channel.

When I shoot in extremely low light, I usually just let ISO 1600 under-expose, as it must.  If I don't like the lens wide open, I'll use manual; if it's OK open and I don't need more DOF, I'll use Tv-pri.

If there were RAW converters that were more oriented towards removing horizontal and vertical line offsets at greater than 12-bit precision, and binning capabilities, you could really push 1600 quite a bit.  As it stands, I have to do all that stuff manually, and it is cumbersome, but I can get ISO 40,000 images that look halfway decent at 4x6 inches.

[digitaldog]

That's right, it's not "real". But it means that if you want to underexpose with three stops, you can do it (the limit for exposure compensation is +/- 2 Ev on any non-1-series digital; the 1-series have +/- 3 Ev).
[a href=\"index.php?act=findpost&pid=72051\"][{POST_SNAPBACK}][/a]

Exactly. It might not be real but the results are an exposure I can work with in such extreme conditions.

[John Sheehy]

Exactly. It might not be real but the results are an exposure I can work with in such extreme conditions.
[a href=\"index.php?act=findpost&pid=72059\"][{POST_SNAPBACK}][/a]

You can work that way, but there are advantages to working with ISO 1600 instead.  In low light, I typically have my camera set to ISO 1600 in Tv mode, with the slowest shutter speed I feel I can get a majority of non-blurred shots.  I set the EC to +1, so if there is enough light that the camera can actually expose for +1, I have "a better ISO 800 than ISO 800".  If it fails by 1 stop, I have a regular ISO 1600 exposure.  If it fails by two stops, I have an exposure with the same noise characteristics as ISO 3200, but with a stop more highlights.  This works very well if you are using a slow lens that has sufficient DOF and is fairly sharp, wide-open.

[Ray]

You can work that way, but there are advantages to working with ISO 1600 instead.  In low light, I typically have my camera set to ISO 1600 in Tv mode, with the slowest shutter speed I feel I can get a majority of non-blurred shots.  I set the EC to +1, so if there is enough light that the camera can actually expose for +1, I have "a better ISO 800 than ISO 800".  If it fails by 1 stop, I have a regular ISO 1600 exposure.  If it fails by two stops, I have an exposure with the same noise characteristics as ISO 3200, but with a stop more highlights.  This works very well if you are using a slow lens that has sufficient DOF and is fairly sharp, wide-open.
[a href=\"index.php?act=findpost&pid=72107\"][{POST_SNAPBACK}][/a]

I agree. Basically, if you expose an image at ISO 1600 fully to the right, then at ISO 3200, with the same exposure, the highlights would be blown. If you underexpose by 1 stop at ISO 1600, because there are extreme highlights you want to preserve, then at the same exposure ISO 3200 would get you the same result. Why bother with these differentials? Just use ISO 1600 so you know where you stand, so to speak.

[jani]

I agree. Basically, if you expose an image at ISO 1600 fully to the right, then at ISO 3200, with the same exposure, the highlights would be blown. If you underexpose by 1 stop at ISO 1600, because there are extreme highlights you want to preserve, then at the same exposure ISO 3200 would get you the same result. Why bother with these differentials? Just use ISO 1600 so you know where you stand, so to speak.

I'm repeating myself, but anyway:

Because you need the extra underexposure.

The example I cited wasn't about needing -1 Ev, it was about situations where you might need e.g. a shutter speed similar to -3 Ev without entering manual mode. Or in the case of a 1-series camera, -4 Ev.

I know that the answer is "buy the Nikon D2X(s) instead, you doofus" because it offers a more sensible +/- 5 Ev compensation.

[Ray]

The example I cited wasn't about needing -1 Ev, it was about situations where you might need e.g. a shutter speed similar to -3 Ev without entering manual mode. Or in the case of a 1-series camera, -4 Ev.

[a href=\"index.php?act=findpost&pid=72118\"][{POST_SNAPBACK}][/a]

Jani,
Perhaps you need to expand on that a bit. First, I've never had reason to use -3 EV at ISO 1600 (no argument in itself), but if I did want to, I'd use manual mode. Is that difficult? You can see the exposure reading change in the viewfinder as you turn the exposure wheel. When you get down to -2 EV, whatever the exposure reading is, halve it and keep turning the wheel till you get your -3 EV.

Are you saying there's a quicker and easier way using ISO 3200? Maybe I'm not feeling very bright at the moment so be patient. Is your method to make a note of the -2 EV exposure reading at ISO 3200, then switch to manual mode, change to ISO 1600 and use that same exposure, thus saving yourself the complicated task of dividing a number by 2?

[digitaldog]

You can work that way, but there are advantages to working with ISO 1600 instead.  In low light, I typically have my camera set to ISO 1600 in Tv mode, with the slowest shutter speed I feel I can get a majority of non-blurred shots.  I set the EC to +1, so if there is enough light that the camera can actually expose for +1, I have "a better ISO 800 than ISO 800".  If it fails by 1 stop, I have a regular ISO 1600 exposure.  If it fails by two stops, I have an exposure with the same noise characteristics as ISO 3200, but with a stop more highlights.  This works very well if you are using a slow lens that has sufficient DOF and is fairly sharp, wide-open.
[a href=\"index.php?act=findpost&pid=72107\"][{POST_SNAPBACK}][/a]

I'll give that a try, thanks.

[jani]

Perhaps you need to expand on that a bit. First, I've never had reason to use -3 EV at ISO 1600 (no argument in itself), but if I did want to, I'd use manual mode. Is that difficult?

It's not difficult, but it's a different operation, since it would also involve switching modes. Since the camera remembers your last manual setting, that may be wildly different from your e.g. Av setting, so getting back to the correct aperture and fiddling about can cost precious time.

Switching ISO level on the 20D, 30D or 5D is, by comparison, really fast.

Are you saying there's a quicker and easier way using ISO 3200? Maybe I'm not feeling very bright at the moment so be patient. Is your method to make a note of the -2 EV exposure reading at ISO 3200, then switch to manual mode, change to ISO 1600 and use that same exposure, thus saving yourself the complicated task of dividing a number by 2?

No, I'd just set it to 3200 and take the shot, if it was that important to get it.

I realise that this is of very limited use; it's just the cases where you need to get there quickly with a minimum of hassle, and you're still running in either program or aperture priority mode.

If I'm in manual mode in the first place, then locking the exposure is just the way to do it.