I use Adobe's DNG converter to convert to DNG, then use Aperture. Haven't seen any such noise with this workflow, 50D is a bit worse than 1D3, which is no surprise...
shrink1 wrote:
I use Adobe's DNG converter to convert to DNG, then use Aperture. Haven't seen any such noise with this workflow, 50D is a bit worse than 1D3, which is no surprise...
I agree. I would put the noise somewhere in between the 40D and 1D3. Unfortunately, I sold my 1D3 before picking up the 50D so I can't do any direct comparisons, but going by samples I still have from the 1D3, the 50D is not as good at high ISO but pretty close at ISO 800 which is my primary concern.
Jim Victory wrote:
It is definitely ACR. Try converting them in DPP and you'll see the 50D is better. This is why I don't use ACR. I have noticed this in every Canon camera I have owned which includes all of the xxD series and 1 series.
Also it looks like the 24-105 needs a little MA.
Jim
ACR or LR only make the nosie look worse if you use the crappy default sharpening values, which really are of almost no use. If you have a noisey shot then make sure you use a high mask value, a small radius and low detail and you can use a large amount value, the mask will ensure noise in large open areas is not accentuated.
Combined with a luminance value of 10-25 it does very well.
I typically use
amount 60-90
radius 0.5-1.0 depending on detail size and camera (0.5-0.7 for 5D, 0.8-1.1 1D II)
detail 20-80, 20 for noisey images, 80 for clean images with a lot of fine detail
mask 25-80
Even if the image is not too noisey, mask should be used as it will avoid sharpening bullred backgrounds or the sky. Hold the alt key down on all the sliders in turn to see what's happening. Also since detail sliders acts like USM, it'll create halos, but if you use a mask value of say 35+ that is avoided, and allows for larger amounts to be used.
Avoiding sharpening at all in ACR/LR isn't wise as it seems as though detail extraction is linked to this process. ie if you have no sharpening in ACR/LR you can't seem to get the detail in PS, unlike with say C1, which extracts the detail irrespective of sharpening.
Until they updated ACR and LR to the new sharpening algorithms I agree they were hopeless for extracting detail and making images look noisey. This is no longer true and ACR matches C1 for detail extraction. it also means DPP is consigned to the waste bin for me.
It is obvious that both shots are not taken from exactly the same position, and the shot on the right shows greater magnification, so why wouldn't the noise also be magnified compared to the 40D shot
Imagemaster wrote:
It is obvious that both shots are not taken from exactly the same position, and the shot on the right shows greater magnification, so why wouldn't the noise also be magnified compared to the 40D shot
Actually these shots were taken from the same location. And although the 50D shot had greater magnification at the time of shooting, I uprezzed the 40D shot to 15mp's. But I really don't want to end up in a struggle with you over this.
I am genuinly interested in the 50D. I have the money burning in my pockets for three weeks already. But ever since I saw the "crappy" ACR 4.6 beta conversions I have my doubts, since I use ACR/LR exclusively for RAW conversions. I don't want to shoot JPEG or use DPP in order to get acceptable results.
So, if ACR 4.6 beta can produce better results with different settings or better pictures, all the better. It could also be that the 50D IR used was a dud. Or whatever reasons more... So to rule out any of this, maybe you or anybody else can provide me with a couple of properly focused/exposed high ISO 50D RAW files so I can see if that will produce better results in ACR 4.6 beta...
csd2020 wrote:
ACR 4.6 RC works quite well if you are skilled with Noise Ninja or some other noise reduction software. Very easy to limit luminance NR while going strong with chrominance NR to get very satisfactory results. Sorry for the large pics but here are two 100% crops, 40D first. Shot at identical settings, ISO800. Same settings applied to both with no NR converted in ACR. The 50D crop is downsized to match the 40D size.
Are these sample shots treated with NN or other similair noise reduction software?
ISO 3200. Neutral picture style. No sharpening. Noise reduction is set to off in both cases. (note: with DPP v3.5, EOS 50D defaults to heavy NR, 40D defaults to no NR). EOS 50D was downsized to 10 MP, and compared to the EOS 40D.
In terms of high ISO noise, I see very little practical difference between the two. The EOS 50D looks slightly noisier, but that may be attributed to the slightly different shooting conditions.
ISO 3200. Neutral picture style. No sharpening. Noise reduction is set to off in both cases. (note: with DPP v3.5, EOS 50D defaults to heavy NR, 40D defaults to no NR). EOS 50D was downsized to 10 MP, and compared to the EOS 40D.
In terms of high ISO noise, I see very little practical difference between the two. The EOS 50D looks slightly noisier, but that may be attributed to the slightly different shooting conditions.
Thanks for posting this DPP comparison... I have seen similair posts and heard 50D users claim the 50D RAW files are much cleaner when converted by DPP 3.5. And based on what I have seen so far, I agree with that. However, I don't like what DPP is doing to details (as opposed to ACR/LR) and I absolutely hate DPP's interface. No way I am going to build a workflow around DPP. I don't understand though why ACR 4.6 does such a poor job in the noise department. Well, obviously it is beta and the 50D isn't even officially supported by it. But maybe I am overlooking something else here, or maybe the IR RAW files just suck. You tell me... In any case, I wont be buying a 50D just yet.
Again two Imaging Resource RAW images with better quality (ISO 1600). I converted them with ACR 4.6 beta with no sharpening/NR applied. I uprezzed the 40D files to 15MP. These are 100% crops:
Just like I saw in another thread where a guy was comparing his 50D to his 1D3 and DsMkII, the 50D seems to be pretty noisy. I'm guessing it's a by-product of the larger more densely packed sensor.
