camera nerdery

09 Nov 2009

Okay, some serious camera gear nerding below.. read at your own peril:

So, I have been having some technical difficulties with my 20D, and it’s a fair bet that it’s going to die sooner or later – and probably sooner. While I’d planned on using it till it died, it’s dying a slow malfunctioning death rather than a quick, painless one. So I’m starting to consider options of replacement sooner than later. I am nearly 100% satisfied with my 20D as it is, and I’d consider replacing it with another newer 20D, were it not for one sticking point: the noise. I’m not happy with the performance of the 20D at ISO 3200. More on that below.

I’ve narrowed it down to a few scenarios, but it’s raising some interesting questions:

Scenario One: Canon 7D

This is a pretty tempting option, but it ain’t cheap. $1700 is a lot of money. There are all sorts of various bells and whistles that come with this newer line of camera, but the primary draw for me is the advertised ISO capabilities and the noise in the samples I’ve seen so far. I do a lot of available light photography, and a lot of it in low light. I abhor a flash, but with the recent addition of the Canon EF 50/1.2 to my kit, I’m starting to push the limits of optics in getting better pictures with less light. The natural next step, then, would be to work on the camera end. So, pretty tempting here.</p>

**Total cost to me: $1700</p>

Major downside: uh, $1700 total cost to me.

Major upside: I can take pictures in the dark now?</b> </dd>

Scenario two: Canon 5D (mk I)

The other lingering option I’ve always had is to go full-frame. This is tempting for a number of reasons: narrower lenses become wider, the full frame sensor is Very Pretty, and so on. The problem here is that the Canon 5D is EF mount lens only – meaning that EF-S lenses do not work. And what’s one of my favorite lenses in the world? My Canon EF-S 10-22 – my go-to (super) wide angle lens. So, I had previously been stymied by the fact that if I moved to a 5D, I’d also have to sell and replace the 10-22 – the nearest equivalent in Canon’s line for a full-frame sensor being the 14mm prime, or the 16-35 zoom. Both of are basically prohibitively expensive, at least when coupled with a concurrent camera upgrade.</p>

What I hadn’t considered, though, is that on a crop sensor (i.e. on my 20D) that 10mm end of the 10-22 is basically 16mm. It occurred to me that I actually have in my possession, right now, a Sigma 20mm/1.8 lens. This lens, at 20mm, on the full-frame sensor of the 5D would not be that different from the 10mm (16mm) on the 20D. Despite the fact that I sorta reamed the Sigma 20mm in a review, it was entirely because it didn’t really suit my intended purposes – a “fast” wider lens for low-light portraits, etc. It’d probably still make a splendid landscape type lens to replace the 10-22.

So, suddenly a full-frame camera isn’t seeming so far-fetched, especially because I could sell the 10-22 and pay for most of a 5D with it. I even know someone selling one that I know is well-cared for and a good deal.

**Total cost to me: $150-200</p>

Major downside: the Sigma 20/1.8 on the 5D still wouldn’t be as wide as my 10-22 is, and (probably?) not as sharp. Also, the 5D is a really heavy camera. that plus the 50/1.2 would be a brick on my back.

Major upside: beautiful 35.8 x 23.9 mm full-frame sensor. better noise performance than my 20D, but not as good as the 7D (maybe? See below)
</b> </dd> </dl>

So, the major sticking point for me really comes down to the low-light/noise performance of the 5D’s full-frame sensor versus the far newer, smaller APS-C sensor. Although by the numbers alone, the clear “winner” is the 7D: witness this image (warning: 8MB JPG) – a sample shot taken with the Canon 7D at ISO 12800.

But numbers aren’t everything. The 5D’s sensor is a different beast. I was curious to see what comparisons look like between the Canon 5D and the Canon 7D look like at 3200 – the highest common ISO they share. From my cursory poking around, they don’t really look that different, but it’s not conclusive.

Further, based on what I know about digital camera sensor ISO (a murky understanding at best), it gets complicated quickly. (To know what I know – or at least what I’ve read, try this article and the followup. It’s dense reading, and if I read them again, I’d probably find the answers to my question/supposition, but blog posts are more fun.)

Here’s my thinking – anyone that knows more about camera sensor ISO stuff should chime in and correct me if this is wrong. A camera’s sensor has a certain dynamic range that it can capture and store, measured in bits per color per pixel. A JPG, for example, stores 8 bits per color per pixel, whereas the RAW data that is stored from your average DSLR is 12, and in newer sensors, up to 14. So, there’s a lot more dynamic range being captured than is typically condensed into a resulting JPG (one of the advantages in shooting RAW). So, based on that, I’m guessing that the resulting image you can get from a picture taken at ISO 1600 doesn’t very that much from a picture taken from the “low end” (read: boosted exposure) of a shot taken at ISO 3200 – or at the very least, that there’s a lot of shared “overlap” in the dynamic range data stored for both. My experience doesn’t bear this theory out 100% – pictures I take at 1600 seem marginally better, even when underexposed and boosted to “ISO 3200” levels – in particular when it comes to my “noise grid” issues.

So, my question is: if I took an underexposed shot with the Canon 5D at ISO 3200 and bumped up the exposure from the RAW image, would it even be remotely comparable to the noise/quality of a shot at a similar exposure level with the 7D taken at ISO “6400” or “12800”? (That is: how much of ISO “6400” is really just re-centering the exposure around dynamic range that is already somewhat present in the shot at ISO 3200 – i.e. cheating.) My guess is probably a qualified “no” – where the qualification is “but maybe good enough”. That is, it may be good enough that it’s not worth dropping an additional $1400 for the 7D.

Cameras are complicated. If I went the 5D route, would I still basically have the same noise limitations – but just on a bigger sensor? Or would it be noticeably better than my 20D? Thoughts?

UPDATE: My friend Chris pointed me at this site, which has comparisons of various cameras’ performance, including “low light” performance. Details on how they measure it here, but the tl;dr of it is: the highest ISO that the camera maintains ‘sufficient’ image quality. It doesn’t have any data on the Canon 7D yet, but for some other comparisons:

Canon 5D: ISO 1368
Canon 20D: ISO 721
Canon 40D: ISO 703
Canon 50D: ISO 696

These are pretty stunning to me – that the 5D would be so much better. I guess the larger photosites on a much larger sensor do carry a considerably significant advantage. If the 7D’s numbers are similar to the 50D, that’s a pretty big check next to the 5D option – hell, even if the 7D matched the 5D’s numbers, the 5D’s cost advantage is still a huge win. Although, *technically*, just because the “acceptable” quality dropoff is higher for the 5D, that doesn’t necessarily mean the rate of quality deterioration isn’t higher for the 5D than the 7D – i.e. the 5D might be better or equal to the 7D at ISO 1368, but it might go to complete shit quickly above that, making the 7D a clear winner at ISO 3200+. Phew, complicated. It’d be easier to say once that site has numbers for the 7D.