Hey Hawk, it certainly is a *lot* to think about, and can be overwhelming without a solid technical understanding to begin with--that was a pretty advanced article, so it's totally understandable that it is a little confusing; less than a year ago it certainly would have confused me as well. I shoot events and science magazine stuff as well as some freelance PJ with dual 7D1s and 17-5/2.8, 70-200/2.8L IS, and a fast prime or two, recently upgraded from dual 550Ds (virtually the same as your 600D) so we are in pretty similar spots in terms of gear.
Eventually I will upgrade to dual 5DIIIs and a 24-70/2.8L, but since most of what I shoot is destined for the web or magazine-format print, I've been able to get away with the limitations of crop format thus far, and picked up the 7Ds (one with only 15K shutter actuations) for a phenomenal price on Adorama ($400 for one and $500 for the other), as opposed to the ~$3000 you'd pay for an equivalently capable full frame camera and lens upgrade (5DIII plus a 24-70/2.8L). But if I was shooting landscapes, portraits, still-life, etc., then a single 6D (plus lens) would be a much more useful upgrade then two 7D1s, and likely be in better condition while not costing much more.
Clark has a good deal of interesting and highly technical material, and certainly delves deeper than just following the "conventional wisdom." And I do note, he doesn't just "think" things are a certain way, he demonstrates through reasoning how he arrived at that conclusion. However, it really helps to have a strong background in the technical side of photography before reading advanced articles like that on his site, as there are several caveats to the statement "image noise is independent of sensor size" that he expects the reader to be aware of, and without awareness of which can lead to confusion and misleading conclusions. Most notably that his examples, analysis, and conclusions only really hold true for "focal length limited" photography, where you cannot realistically get closer to your subject (and consequently make it fill more of the larger full-frame image circle).
So, this certainly matters if you are interested mostly or solely in wildlife/bird photography, astrophotography, some types of outdoor open-field sports, surveillance, or anything that involves the subject being very far away from the camera, without the possibility of getting any closer. Let's assume we are comparing a 20 MP 6D to a 20 MP 7D2/70D (and roughly equivalent to your current 600D and the 7D1), both with your 70-200/4L IS (therefore the 6D has much larger pixels, but a lower density of pixels on the sensor). In cases like this, let's say your subject is smaller than your desired framing at 200mm (your longest focal length) on your 7D, you would be "focal length limited;" and your subject would cover fewer sensor pixels on your 6D as opposed to your 7D at that focal length. Therefore, cropped to a similar subject size relative to the frame, your 6D shot would be cleaner, but less detailed then your 7D shot, while if you then resized the 7D shot to the lower resolution of the cropped 6D shot, you'd get essentially the same quality image in both noise and resolution (all other factors besides pixel area being equal).
That being true, this analysis does *not* hold presuming the subject is within a "normal" distance range for most photography, and you can get closer if need be. Let's say you've zoomed the lens to 100m, which results in an appropriately-sized framing of your subject on your 7D's sensor. However, due to the field of view crop of your 7D relative to your full-frame 6D, on the latter the subject will be smaller relative to the size of your frame, resulting in fewer pixels covering the subject, and thus the same scenario as before. However, the important difference is this--here you have the option of either moving closer to your subject, or zooming your 70-200 to 160mm, which would result in an equivalent framing of your subject as the 7D (due to that latter's 1.6x field of view crop) and the subject will cover the same number of pixels on that sensor. However, the 6D's pixels are of course larger, which results in nearly 2 stops better (photon) noise performance (all other factors being equal) at the same resolution.
Another way to look at it is that your 70-200/4L IS captures the same amount of light either way, but the larger sensor collects much more of that light that is "wasted" on the smaller sensor. Therefore, in non-focal length limited cases like this scenario, which can be 95%+ of many (or even most) types of photography, full-frame does offer a significant advantage in low light with the same lens (assuming it is full frame compatible, of course), equivalent in noise and depth of field to using a lens around 1.5 stops faster on a crop sensor (like upgrading a f/2.8 zoom to a f/1.8 prime, or a variable aperture kit lens at a middle focal length to a f/2.8 zoom.)
Hawk said:
Seems like overall aperture through lens is more important than F-stop overall.
Not quite sure what you are trying to say here; I think you mean the absolute aperture diameter of the lens (say 25mm) as opposed to its f-ratio (the focal length of the lens divided by that diameter, say f/2 for a 50mm lens with a 25mm aperture diameter). Which one is more critical depends on the context, so it's meaningless to broadly state that one is more important than the other. Don't worry about this for now; it's not vital to understanding the subject at this level.
Hawk said:
And that ISO isn't anything but post gain.
Again, this is technically true. However, to understand the real significance and complexities of this requires some background in how sensors work, particularly where and how that gain is introduced. I'll try to explain it as simply as I can. The key thing to understand is that there are two sources of noise in a photograph, shot (or photon) noise and read noise, differentiated by where the noise is introduced into the image. For purposes of low-light performance, shot noise is the more important, and in isolation depends strictly upon the amount of light hitting the sensor over the finite length of your exposure. It creeps in because light is transmitted by particles called photons, which each carry a discrete amount of energy that is detected by your camera's sensor, and have an element of randomness in their arrival. At high levels of light, where thousands or millions of photons are hitting each pixel, a few more or less aren't going to make that much difference. However, at low light levels, only a small number of photos may arrive at a single pixel over the time of your exposure, and so if one or two fewer hits one pixel as opposed to an adjacent one due to this randomness, this random fluctuation is observable in the final photo as noise.
While sensors of higher efficiency are theoretically possible (ie fewer photons are "lost" somewhere between the lens and the detector), we are already nearing the practical limits of this, so the main ways to decrease shot noise are either to have a longer exposure time (shutter speed) allowing more photos to be collected (but risking motion blur), a wider aperture lens to gather more light for the sensor (more expensive) or making the pixels larger, either on the sensor (each gather more light due to their larger area) or downsizing images in post, "averaging out" image noise over multiple pixels. If this were the only source of noise, there would be no need to set an ISO on your camera, and you could merely digitally "gain up" in post to brighten the image to your satisfaction, with no more noise than you would have gotten amplifying the signal on the camera. Some of the very newest sensors approach this property, which is called "ISO Independence," such as the Sony sensors Nikon uses in the D800/10 and D750. With these sensors, ISO 200 and above, there is minimal difference in noise whether taking a photo at a higher ISO to exposure properly or shooting at ISO 200 regardless and raising the exposure of the RAW file by the same number of stops in your RAW converter.
Unfortunately, for most other sensors a major consideration is read noise, which is introduced by the camera's electronics after the signal from the sensor is amplified, but before/as the analog data is converted to digital. This mainly affects darker areas of the image at low ISO, particularly if you try to brighten them in your RAW converter. Read noise stays at a constant level relative to the final signal strength before digital conversion, so if you amplify the signal before you digitize (raise the ISO), you amplify the shot noise along with the good signal but not the read noise, and thus you raise more of the useful signal above the constant low-level read noise.
Of course, if you amplify it enough so that all the useful signal is above the read noise level, you gain no better quality since you are still limited by shot noise, and risk bright areas of the image being amplified all the way to white (which cannot be recovered in post). On most Canons, this limit falls around ISO 3200 (at least for crop frame) and on the best Nikons, as low as ISO 200. However, again, in low-light you are mainly limited by shot noise, since after amplification the signal is sufficiently noisy that read noise is a negligible factor. And, as stated above, larger pixels on a larger sensor (or wider aperture, or longer shutter speed, with their respective tradeoffs) will increase the amount of light you collect at the source, raising the signal further above shot noise.
Hawk said:
I do notice how much better my 70-200 was than my 17-55.
If by this you mean the images from the 70-200/4L IS look sharper/etc. than those from your 17-55/2.8, this likely has little to do with what Clark discussed (unless you are shooting the same subject at the same distance with both, and then cropping the 17-55's image to the same subject size, which would be absurd). Rather, the 70-200/4L IS is a very sharp lens, significantly sharper across the field on crop frame at every focal length then the 17-55 at its sharpest, as well as with less distortion, vingetting, etc. both due to its more modern and high-quality construction, and the fact that since you get better images from the center of a lens than its corners, you are only using best portion of the image on your 600D from the full-frame 70-200 (as opposed to the crop-frame 17-55).
Furthermore, many would argue that the subjective qualities (bokeh, contrast, color rendition) of the 70-200 are superior, and due to the longer focal length you get greater subject isolation and more background blur, which many find to be more pleasing. Finally, there is the fact that lenses generally perform better at smaller apertures, and since the 70-200/4's max aperture is narrower than the 17-55's, you would likely be using it closer to it's peak performance than the 17-55, at least some of the time. Of course, "better" could mean any number of things, so describing one lens as definitively "better" than another without qualification or context can, again, be somewhat meaningless.
Hawk said:
He seems to think 200MM F4 is 50MM where as 55 F4 is 13.5MM...
I assume you are discussing the actual/physical aperture diameter, in which case your first statement is certainly true. Again, he doesn't "seem to think" this, it is objective fact. You can easily verify this for yourself. As you probably know, your f-number is the ratio of a lenses' physical aperture diameter to its focal length, so therefore f(focal length) / f-number = Aperture diameter (That is why aperture is given as f/[f-number], as in f/4). So 200mm / 4 = 50mm, and 55mm / 4 = 13.5mm.
Hawk said:
...so more size less noise. Lets in more light to sensor.
This is true...but ONLY for a given focal length, so this would not hold for the example you gave. Optics dictates that a lens of 4 times the focal length would need four times the light (4x the aperture diameter) to produce an image of equivalent brightness on the focal plane (the sensor). Therefore, both lenses would produce images that are of roughly equal brightness, as we would expect given their equal f-ratios. That is one reason why we use f-ratio to compare the apertures of different lenses, rather than the raw aperture diameters. It is interesting to note, however, that increasing aperture diameter with focal length to maintain the same f-number (brightness) is the reason longer focal length lenses with the same f-number have shallower DoF, since DoF is determined by the aperture diameter, not (directly) by the f-ratio.
Hawk said:
He thinks pixel pitch and overall light plays alot more with less noise than just FF vs APSC debate.
Not exactly, as I've tried to explain above, but he does make the point that total sensor size is not the only thing that matters. At any given resolution, pixel pitch (and thus light-gathering area per pixel) is directly related to sensor size, and with the same framing, a FF sensor of a similar generation should always produce a cleaner image at the same resolution. Once the desired framing is tighter than that projected by the lens on the smaller sensor (focal length limited), then a full-frame sensor of the same pixel density (ie same size pixels) will produce a final cropped image no better than that of an equivalent APS-C sensor with the same lens.
However, for many types of photography, these situations are rare; it all depends on what sort of photography you do. Furthermore, this assumes a perfectly resolving lens above the diffraction limit, both of which further tilt any advantage away from a higher density APS-C sensor, since our modern sensors are at least partially limited by what the lens can resolve, preventing higher-density crop sensors from having as much of an advantage in FLL situations since the lens often can't resolve the extra detail anyway.
Sorry for the huge infodump, but hopefully it helps clear up some of your understanding on this issue! As for what camera to pick, this really depends on what type of photography you plan to use it for, as described above. Of course, if you have questions or still don't understand something, please don't hesitate to ask.
Good luck!
C. A. M. Gerlach