I don't agree. The less perceived noise is due to the larger sensor area not more pixels. If it were due to more pixels this would say Canon could magically improve the 10d by increasing nothing but it's pixel count to 9MP.
Absolutely! You agree with me, it's the larger sensor area not having to be enlarged as much that causes the 1Ds to have less practical noise. You might have a problem with the low pass filter effects in each different case (digital resampling vs. printing) but they are very closely related.
I don't agree, but I think we will have to agree to disagree. One thing 10D owners forget is they have rediculously low headroom in the RAW file. It's basically 1/3rd stop if that. TThe 1Ds is closer to the Kodak 14n, and thereby decreases it's apparent SNR by keeping more detail beyond the JPG clipping point which decreases the SNR. It's not Phil's style at all to measure SNR using this advantage (ala the "expose to the right") LL article though.
Although I am a fan of larger sensors, it's worth noting that the current P&S camera's lenses resolve well more than this at large apertures.
I guess you're talking about the total lines resolved here right?
There are times when this isn't even true. From everything I have seen the 17-40 on the 1Ds vs. the 12-24 on the D100 is a good example. Not a fair test without the same sensor (film) obivously.
--
If the 1Ds had only 6MP on the FF sensor and maintained its current per pixel noise performance there will be no difference whatsoever between its image quality and that of the 10D noisewise. The reason is just what Phil said in the quote you wrote. The camera that produces a larger image can stand a somewhat worse noise performance. If you want to create a 1MP image from both the 10D and 1DS to put on a website, you are compressing the 1DS image more and with it you are compressing the noise more, resulting less percieved noise. If you are enlarging to make a print and making say a 25 MP image for a large print, you are stretching the 10D image more than the 1DS image and with it you are enlarging the noise more, again leading to less percieved noise for the 1DS image. Yes the percieved noise from a 9MP sensor image (enlarged or decreased in size) will be better than a 6MP one from the same format IF the per pixel noise is similar or just slightly worse.
The analogy I made to film is simliar to what I just said but you have to keep one distinction in mind. Film formats of a certian brand and type have the same photosensitive elements per unit area of the film frame. Sensors of different formats may or may not have the same image capturing elements (pixels) per unit area fo the sensor. In the case of the 10D and 1DS, not. You can not compare the differences between electronic sensor formats the same way you do with film. It is the relative sizes of the final image that plays a role in the percieved noise
But I agree with that
Why do people keep claiming that the Nikon mount is not practical for a full frame sensor? The image circle is fine for film. And both Kodak and Horseman (Digiflex) have proven the F mount works with a full frame sensor.
Yes, but I believe this is an artificial constraint, and didn't apply to my example. I guess we'll have to wait to see how the next 1Ds performs. I'm quite sure it will be superior in the noise area to the current 1Ds. I am realistic though, I do know we will not have perfectly clean ISO 6400s one day, since it's beyond the physics limits.
I just have one nitpicky comment involving this. It's not that the "perceived noise" goes down when downsampling the 1Ds to 1MP, the SNR really does increase. This is due to the low pass filtering of the higher frequency components of the noise which is required so you don't end up with a jagged, aliased 1MP image. This effect is a bit more complex when printing (both hardware and software effects) however the result is very much the same.
I have been arguing against that very statement since my first post. Larger pixels equal better pixel level noise for differnt formats is simply not necessarily true and the example the 1Ds and 10D just poves that.
We are back to what I was saying in my first post. If we can build an 11.5 MP sensor and assuming that I can define something called "same technology" , will that same technology give me 29.4 MP with the same sensitivity as the smaller sensor and not much worse per pixels noise? We do not have teh slightest clue. The 10D and 1DS example says that at least the senstivity of the smaller sensor will be much better. We also do not know if per pixel noise will vary linearly with pixel count. That if the pixel level noise performace of a 29.5 MP FF sensor relative to the pixel leve noise performace of an 11.5 MP 1.6x sensor be the same as the relative pixel level noise performance between 11.5MP FF and 6MP FF. It might turn out that things get worse faster with increasting the pixel count for one of the formats.
Once you put the "common wisdom" generalization about pixel size and noise aside, you will realize there is no sufficient knowledge at all.
I am waiting for this to get verified myself. We will have to wait till Nikon can make (if they will) a DX format sensor with around 11.5 MP to see the advantage of digital lenses or of the central crop of the image circle.
The AA filter is before there is any noise. If you're saying the 10d can get away** with more smoothing due to the lack of much high frequency detail due to the previous filtering of the data, then I'd agree that's possible.
The way I inderstand this is that you are saying that the AA filter is a digital one. My understanding is that AA filters in Digital cameras are are all optical
No, this isn't what I was getting at. AA filters need to be analog, so they block the frequencies beyond nyquist before you sample. If you do this digital, you have to guess what is aliased information and what is not. Beyond this, if you've bayered, the info gets spread out and turns ugly.
Yes, although there are exceptions eg. 14n.
--
I'm sure this is how it's done and this is the secret behind D60's/10D's "silky smooth" images. That's why 10D gets away with better noise suppression than 1Ds - not because its sensor is inherently better in noise, but because stronger AA allows the software to extract the high-frequency info over the blurred real data and remove this noise.
If I had to guess, the 1Ds is based on the same technology as the D60, and it provides BETTER noise per pixel than the D60. The 10D was designed with some process refinements. I believe you will see a new 1Ds in the next 12-18 months that beats the 10D in noise per pixel.
This is not "common wisdom" but engineering fact. You keep bringing other variables into this, which takes the controlled element away from such a comparison.
