7d, IS0 4000
I didn't even realise I'd bumped it up this high - I thought I was at 3200. Sure, there's some noise but... really, I can't complain! They were only lit with some blacklights and spill from the stage (hence why I converted it to BW - the mixed light colours were really whacky). Focus was on the center guy.
50mm, f2.2, 1/50, ISO 4000
50mm, f2.2, 1/50, ISO 4000
facebook | photo site |
0
Comments
I just know that people are often still questioning the 7d's ISO capabilities. I'm sure a 5dII or d700 is better yet, but really, for a crop sensor - and with that many pixels! - I find this fairly astonishing.
I looked at the original (size) of this image on your site, and I'd hate to see what that would look like printed even at 8x10.
I own a 7D, but put me down as one who's not completely sold on it's high ISO noise levels. I did a little testing between my 7D & old 30D. These things are subjective, but to my eye, at ISO 3200 the 7D seems to only be one stop cleaner than the 30D. In other words; the 7D ISO 3200 shot looked just about like the 30D ISO 1600 noise level. Not the quantum leap that I'd hoped for given 3 generations newer technology.
Not trying to rain on your parade, but I'd hate to think that others might get the impression that the 7D, at high ISO and underexposed shots, (like the one you posted) are going to print out beautifully.
Disclaimer: I've shot with 5D & 5DMkII for some time now, so I've been high ISO/low noise spoiled.
Also - doesn't noise show LESS on prints than in web images? That's what I'd always understood. Not that it matters here since I have no intention of printing it (I'm not even using the shot - it was more one of those
Yeah, GB - unfortunately, with performance photography... no flash allowed!
I agree that there is not a "quantum leap" between the random high-ISO noise levels of the Canon 20D/30D and the Canon 7D, but a combination of lower noise "and" more than twice the pixels (potentially twice the detail) needs to be appreciated for the photographic achievement that it is.
Normally any increase in pixel density, without any other change in sensor technology, will increase random noise at about the same rate as the increase in pixel density itself. What this means is potentially no real increase in image detail. In other words, if you add noise at the same rate as the number of pixels increases, that just means that more pixels are being used to represent noise instead of real detail. This is the primary reason why P&S camera images do not match dSLR image quality even though they can have the same pixel count and very nice lenses; the pixel "density" of the P&S simply does not allow the lower random noise levels of the larger imager in the dSLR and the higher noise of the P&S obscures more image detail.
Follow me so far?
For the 7D to have enhanced control over random sensor noise "and" to have more than twice the pixel density really is quite an achievement. This represents a 2 x 2 improvement (twice the noise control and twice the pixels) that would have "required" a much larger imager using the technology of the 20D/30D imager.
So what does all this mean from a perceptual context? In other words, how much of an improvement is going to be visible in a print?
Because of the rather odd and quirky way in which our eyes see and our brains process that vision, we may see up to approximately twice the image quality, at higher ISOs and with larger print sizes, in the 7D print vs the 20D/30D print. Better lenses are part of the equation too because image detail may not be present in lesser lenses so the 7D camera will not make use of the extra resolution unless the lens can provide the extra image quality to begin with.
Lower ISO settings will provide less opportunity for image improvement because at lower ISOs random noise differences between these cameras are less, and the 125% improvement in resolution from the 20D/30D to the 7D will only improve our perception of the image by around 60%, not the doubling that the numbers alone seem to indicate. This is because our eyes and brain process the printed image by its "area" (length "times" width). Also, the 20D/30D imager is already very capable with image enlargements up to 8" x 10"/8" x 12", so prints that size or smaller will not show much difference with most subjects.
Moderator of the Cameras and Accessories forums
Natural selection is responsible for every living thing that exists.
D3s, D500, D5300, and way more glass than the wife knows about.
I hope it don't cause I like his brain, and his ability to share ad-infinitum with the not so well informed! vis: Moi!
Well, after another poster gave me the tip on How to finally get my 5Dmk2 to H1 AND H2, I took some test shots....and to be fare, I did take some with My D700 too...SOOC in the Canon was incredible! After Noiseware, a toss up...
Diva, I had no idea that these things could make usable images that high!
The Canon @H1 was superb! eye's clear, etc.
I need to do more testing....because thus far, I'm leaning more toward the Canon!
Oh, but both were shot with a Nikon lens: 100MF f/2.5 AIS. @ f8, natch light.
He is what I do understand with the caveat it may be completely wrong.
A sensor is just a piece of silicon with a bunch of CMOS transistor in it. Initially it all the CMOS transistors are "turned off" but as light falls on the silicon more and more of them turn on based on some random distribution. Also some of them turn on or off spontaneously. Together this causes noise, because some of the transistors end up in the wrong state.
Once the shutter closes the circuity in the camera counts the number of CMOS transistors that are turned on in each pixel. The internal wiring in the sensor groups CMOS transistors into groups, and each group is a pixel.
So the sensor acts sort of like a CMOS memory chip except that the transistor sare turned on by light rather than by an signal internal to the chip.
The thing I don't understand is the total number of CMOS transistor that end up in the wrong state had nothing to do with the number of pixels in the sensor. A pixel is just a group of CMOS transistors wired up together for reading purposes. The sensor temperature, stray light, and circuit noise cause the CMOS transistor to flip the wrong way and it doesn't matter what pixel they are in.
Soooo, given two sensors of the same size, technology, etc they would have the same total number of CMOS transistors. The one with higher pixel density would have fewer CMOS transistors per pixel. I would expect the same amount of noise in the final image to be the same, because the percentage of "wrong state" CMOS transistors per pixel would be the same.
If the pixels on one sensor had twice the area of the other, it would cover twice as many "wrong state" CMOS transistors... that's why the percentage of wrong state transistors would be the same.
It's hard to test this out because I have only one camera, and even if I picked up a second I don't know where I could find one that differed only in the number of pixels.
This is just of interest to me. Ziggy if you have some more details or can point me to something that digs into the details of how sensors work that would be appreciated.
Tnx,
Dan
http://www.danalphotos.com
http://www.pluralsight.com
http://twitter.com/d114
Trying to separate any of these underlying issues from the total system in a camera is beyond the abilities of most camera users, so it's best to assume that we should limit ourselves to the "output" efficiencies of each camera and the accumulative effect of the system in total.
Assuming that all other parts of the system in a camera are equal, and looking at just the pixel density as the only variable, it's easy to understand the reason why more pixel density results in more random sensor noise (and conversely lower pixel density results in less random sensor noise).
Lower pixel density tends to equalize the random nature of sensor noise because you are taking a single sample of a larger site. That single data access point is accumulating data from a larger sample and the larger sample is using a natural type of normalization (averaging) of the randomness of the media. The same thing occurs with longer time duration sampling (longer time exposures) and/or more samples normalized together (multiple short exposures combined into a single image). All of these mechanisms and techniques will normalize/average the input data resulting in less randomness of the data, thus less random sensor noise.
When you sample a smaller pixel site there is less opportunity for averaging the data, and a greater likelihood that random noise values will have a greater influence on the measured values at that photosite. You can compensate with either longer exposures and/or multiple exposures combined (when appropriate).
Software noise reduction uses various different algorithms to try to achieve similar results. (Basically the best software tries to separate random "and" predictable noise from image detail, and then the software uses normalization against the noise component.)
Moderator of the Cameras and Accessories forums
I see it now, it's not the average number of number of transistors counted, it's the deviation that swamps the count at low light level. The more pixels there are, the more likely it is that you will find a pixels with a noise count that deviates a lot from the average noise count.
Dan
http://www.danalphotos.com
http://www.pluralsight.com
http://twitter.com/d114
Yes, that's a very succinct way of putting it.
Now you also understand why high pixel density is not always a bad thing. At smaller print sizes, where you are averaging the random noise between neighboring photosites, you are reducing visible noise overall, similar to just having a lower sensor pixel count to begin with. The more important factor is overall sensor efficiency (which is much more difficult to quantify).
Moderator of the Cameras and Accessories forums
Natural selection is responsible for every living thing that exists.
D3s, D500, D5300, and way more glass than the wife knows about.
Ya got me going on this High ISo thing. And after reading so much....stuff others have written Nikon versus Canon, I decided to put my old test pig to the test.
Don't bother lookin for the Canon EXIF, it won't be there....must be because it doesn't record it when you use a non-canon lens?? Also though the Nik exif says f/11, it was f/8! Either way, Both were Shot at H1 ( yes! 12,800 ISO) using Nikon 100mm MF AIS f/2.5 lens @f/8, Shutter speed ranged due to light availabilty. Download, bring into LR2, Crop @ 4x5, save as jpg, upload and link.
Canon:
Nikon: EXIF
I think these both are very usable images should a person have need for that kind of ISO. Also I would imagine that Some curves, and some Noiseware would aid these greatly!
Oh, and both were AWB!! Both in-camera noise reductions were on Standard or normal. No Hi-D-lighting, etc.
It also proves that successfully shooting at high ISO is sooooo dependent on accurate exposure. My shots above were, as is obvious, practically in the dark - there was no way to get more exposure without a tripod.
These are great samples, Tom. I mean, realistically, I'd be happy to know I had clean 6400 in the bag,never mind 12.8k!!! I can imagine that my next body will be a ff, although by the time I have the money it will probably be the 5dIII or IV
MKIII! I will be all over that day!
Just went back and Checked the Canon again, from when I last shot. It was earlier, by about 25 minutes, but the Shutter speed was @160. Nik was @ 50...both I think hand holdable! though DOF was still pretty narrow @ 11' f/8 100mm.
Personally The Canon Shot looks Just over the top SOOC, but after a simple crop, they look amazingly similar! My D700 has never really lived up to my expectations in High ISO realm! But, in context and as a print I think this Nik h1 photo would work. The Canon is IMO, Better! So often we see photos @ High ISO of non-organic elements. That is THE ONLY reason I chose to show my un-plugged-Mug! A Human Face is what we are programmed to distinguish! SO to me, it is the only thing we ought to shoot to really get to the heart of the matter. Dark/light/Noise...judge for yourself:
Link to Gallery here Original Size turned on. Tomorrow, perhaps 25k?
Thanks for the kind reply!
The 7D likes to be shot "to the right". I have nearly 5000 ISO 6400 and 12800 shots from the 7D at this point over the last year of band competitions and night sporting events. With the proper settings, then proper noise removal techniques, 12800 shots can indeed be saved and printed as 8x10s. I have about as many high ISO shots on the MKIII as well, and I really miss that camera.
I believe the sweet spot on the 7D though is ISO 6400.
Here's a shot I took at ISO 3200, Oly E-30, 14-54 2.8