Part 4: Canon 50D sRAW Noise and High ISO test

This sRAW thing is fascinating.

I can't reproduce sRAW results from full RAW files. Not even close. I wish I could though.

On the left is a 50D RAW file at ISO 1600, 200mm f/2.8.
On the right is a 50D sRAW2 file at ISO 3200, 400mm f/5.6 (200mm @ f/2.8 + EF-2x)
Both were cropped to the middle 600 pixels with no resizing necessary.

I thought the higher ISO and EF-2x should have ruined the one on the right, but sRAW overcame that. I'll have to try that test again with smaller apertures to be sure.

It seems to me that sRAW should be valuable to bird photographers who can't get very close to small birds in low light and wonder if they should crop a raw file or use an extender and trade the stop or two of light for higher ISOs. And indoor or night sports photographers where high ISOs and fast shutter speeds are necessary and will either crop use the extenders and don't need huge prints anyway.

sRAW is NOT for saving space on a memory card. The files are extremely bloated for their small resolution. And they don't relieve the buffer much at all. Probably because sRAW is generated from the RAW file, but that would slow the camera down (and it doesn't)?

I now tend to use sRAW1 when at ISO 6400 and sRAW2 when at ISO 12800.

Here's an example of an indoor low light sports photo. sRAW2, ISO 12800, 1/1000, f/2.5, 85mm, cropped not resized. Exposed to the right, corrected and heavily processed in Lightroom.

---

Part 3: Canon 50D sRAW Noise and High ISO Test

I shot three sets of black frames for these comparisons. I went from ISO 1600 to 12800 for RAW, sRAW1, and sRAW2. I cropped all photos to 200x200 pixels then resized 200%.

1st set: exposure +4 (max) in Lightroom.
2nd set: exposure 0 + Lightroom noise reduction at 100 (max).
3rd set: exposure at -1.5 with noise reduction at 100 again.

The first set is totally exaggerated and not very useful.
In the second, it looks like ISO 3200 on the 50D is fine and 6400 might be useful.
The 3rd is a good example of how I try to use high ISOs; overexpose by a stop or two and darken in Lightroom. In this 3rd example I'd say 6400 is fine and 12800 might be useful.

sRAW modes seem to improve the sharpness and decrease the noise (more testing to do). I was expecting to see more of a difference between sRAW1 and sRAW2.

---

Part 2: Canon sRAW Noise and High ISO Test ( 40D reference for the 50D )

The previous test was quick and had a lot of variables I didn't bother to think about. This time I did, so I eliminated what I could.

I'm trying to figure out if the sRAW mode on Canon's recently announced 50D will be useful to me. So I'm testing the sRAW mode on the 40D and here are my next round comparison shots...


I decided to eliminate the resizing variable by changing my focal length so the same number of pixels are covering the same parts of the scene (I just cropped the RAW file to the sRAW dimensions and get the same view). The first pair is underexposed and shows the noise differences well, the middle pair is a normal exposure and shows sharpness differences well, and the 3rd pair doesn't change the focal length so the RAW file had to be re-sized to match the same view.

The previous test proved to me that noise was lower using sRAW. This one agrees. The previous test also made me wonder if sRAW was sharper too. This one seems to agree, but makes me realize that the question is more complicated than that. Yes, 4 sRAW pixels are sharper than 1 RAW pixel. The middle pair shows this very well. But that's not how I'd use the camera. I'd use it like the 3rd pair, and after resizing, they are similar.

This does lead me to a valuable conclusion: I expect that canon's 50D in sRAW mode at 7mp will be sharper, and have significantly less noise than an 8mp 30D. I already expected that, but now I'm convinced the difference will be drastic and worthwhile to me.

That means it's time to upgrade! YESSSS!

Other interesting conclusions: The sRAW file size is typically half the full RAW file size, which is unfortunate because the resolution is 1/4. Also, burst mode did last a little longer. I got about 17 RAW files out at 6.5fps, but could get over 20 sRAW files at 6.5fps and the buffer cleared faster. So sRAW mode isn't so much for when you want smaller files, but for fast action bursts at high ISOs.

Details: 24-70 2.8L at 35mm and 70mm and f/8, 70-200 2.8 L IS at 200mm and f/8, 40D, manual focus with live view, on tripod, using mirror lock up and cable release. RAW and sRAW files processed in Lightroom with everything zeroed. Jpegs at highest quality while cropping and bi-cubic resizing. 40D's high ISO noise reduction was OFF this time.

---

Canon sRAW Noise and High ISO Test ( 40D reference for the 50D )

When Canon announced the 50D, the specs that caught my attention were the sRAW modes. I liked the idea of sRAW mode on the 40D but couldn't manufacture a reason to shoot a 10mp sensor at 2.5mp. It looks like the 50D will shoot it's 15mp sensor at 7mp. For some of the things I shoot that sounds worthwhile. Then I wondered if those sRAW files would be less noisy at high ISO and just had to test it...

At first I shot a variety of exposures of the same scene and cropped the full size files to match the sRAW size. A noise difference didn't jump out at me at all. What did was how much sharper the sRAW files were. It looks like I either had image stabilization on in one shot and off in the other (which I didn't), or I used a much smaller aperture to get more depth of field (which I didn't). I'll have to try this again with a tripod and manual focusing.


Then I shot black frames (lens cap on at normally fast shutter speed) at ISO 1600 and 3200 in RAW and sRAW. That's where I see a huge difference in noise. Here's the ISO 3200 comparison. I see blatant noise reduction in sRAW. The noise in sRAw seems to be all luminance and no chroma too (someone please explain that to me). I also see more noise in the 40D at 3200 than the 30D at 3200 which is not what I expected (or was advertised). And I should mention that I had to beat the heck out of these black frames to get them to look this bad/obvious.


I did this test in 5 minutes. Please leave comments if you have suggestions for improving the method.
Details:
AWB, Manual Mode, ISO 3200 and 1600, 24-70 f/2.8 @ 70mm & f/4.5 on the 40D, 70-200 f/2.8 IS @ 70mm & f/4.5 on the 30D, handheld, autofocus on middle point at closest flower, raw files 'barely processed' in Lightroom, no manipulation (everything zeroed) on the flower shots, for the black frames only exposure slider set to +4, output was full jpg at max quality.
jpgs from sRAW files cropped to 800x800 near middle of frame. 300x300 corner for black frames.
jpgs from RAW files bicubic downsized to match sRAW dimensions (1944x1296) then cropped to 800x800 near middle of frame.
for exaggerating black frame noise: cropped 300x300 bottom right corner (because noise appeared strongest there in the 40D) then up rezed 200%, brightness +50 and contrast +25 to exaggerate for visibility.

I have high hopes for the 50D and its sRAW modes now.

---