[Pavel]

I experimented using Matthews D700 with my Samyang 8 mm fisheye, which is specifically designed for an APS-C sensor under low light, high contrast situation.
There are several issues I wished to examine in this experiment. One had to do with the coverage of the sensor and the other with noise/image quality under moderately low light, high contrast situation. The experiment was not well thought through, so there will be some arm waving.

1) D700 shows in the viewfinder the coverage one is to expect from an APS-C sensor lens. it is a "wire frame" similar to the one you get in a rangefinder for a telephoto. The actual coverage of the Samyang was much larger and covered about 1/2 way the distance between the wire frame and the full image size. So far so good. However, the image quality (resolution) appeared low in this "extra" space.

2) I set the camera to ISO 3200. DXO sensor tests suggest that D700 beats D300 sensor by about 1.5 f-stops in terms of light sensitivity and dynamic range. On the other hand, I heard from a number of photographers that in practice, D700 performed a lot better than that. I therefore hoped that 2 f-stops higher than my usual cut off for a tolerable photo (ISO 800) should be doable. At this point, I should say that under high contrast situations, I usually expose for highlights (underexpose to the point where my highlights are just not blown and i bring up the shadows in software). Sometimes I use GNDs and sometimes HDR/multiple images. In this instance, I have been exposing for highlights. This is tough on image quality, because of the combined effect of the noise from high ISO AND an augmentation of this issue by boosting the shadows in post-processing. As I said, this methods works up to ISO 800 with D300.

At ISO 3200 with D700, the noise in the shadows was slightly greater than with D300 at ISO 800 (not based on any measurement - based on my experience). I found that in order to get a relatively low artifact image, I had to do sufficiently radical noise reduction that the image appeared smeared. Not much, but more so than I usually get with D300 at ISO 800. The difference was not large, but it was noticeable. I would guess that as far as shadows were concerned, DXO - based estimate of about 1.5 f-stop gain from D300 to D700 is about right. Midtones and highlights looked fine at ISO 3200 and required no cleanup, confirming what the happy owners were telling me about shooting at high ISO with D700. I am guessing that 2 factors conspire. High ISO on its own tends to exaggerate noise, but the problem is minor or field irrelevant with full frame sensor if the image is correctly composed and the image is not overly contrasty and does not require too much manipulation. Similarly high contrast photos shot at low ISO are much more resistant to noise. It is the common situation however, that the low light situation contain elements with high contrast and so this relationship between high ISO and contrast is very relevant. Furthermore, dynamic range decreases with increasing ISO.

Combination of noise and general softness in the areas outside of the APS-C frame on D700 makes this are ill suited for placing there subject where resolution matters. Unfortunately, I usually place at least a part of my main subject in that area.

The answer - There is no free lunch. But in some situations, you could get a worthwhile improvement in low light conditions shooting with full frame cameras and APS-C lenses, trading resolution for light sensitivity. This trade may be less painful as the number of pixels seem to grow rapidly in anounced or rumored cameras

[Toad]

Interesting experiment. I suppose a full frame sensor gives you room to crop the APS image to fill the frame - so in that sense, a viable option.

Personally I think the whole high ISO thing is a red herring. We used to feel that ISO 400 was "OK" and ISO 800 was exceptional. Now we feel a camera has to do clean ISO 3200 to be "OK". Much as I dislike color noise in dark images, I can't help think that the high ISO race has become the new "megapixel" race. A few years ago, the more megapixels the better - and now that people have become aware that anything much above 18 Mpx is "coal to Newcastle", high ISO seems to have become the new Holy Grail (sorry for badly mixed metaphors).

I am happy that you have found the sweet spot between resolution and noise - even if it involves significant tradeoffs. If you have a moment, I would like to see a few photos from your experimentation session.

[Pavel]

Just a quickie Rob. I was never impressed with the argument for more pixels. However low noise and high dynamic range is and always was a big deal to me. In many situations tripod is not practical or not possible and yet light is low and contrast very high. A good example is photography in public buildings where tripod or flash is not permitted and furthermore you may wish to freeze the motion. I know that there are various tricks and compromizes to deal with that, but often none is satisfactory. Whether you find that an issue or not depends on the type of photographs you take. I have been always fascinated with light and shadow and until recently, light and shadow where typically a key compositional element in most of my landscape/cityscape photos.

[matthew]

First of all, I did warn Pavel that I was going to post a photo of him, so here it is:



But I could be starting a battle that I'm guaranteed to use – Pavel spent most of the morning with his fisheye, so he must have some real blackmail-quality images of me.

I'm probably one of the few people who doesn't really care about the D700's low-light ability; my interest has always been the tremendously forgiving low-iso raw files. But with that said, I'm less picky about noise and typically prefer smaller prints than Pavel does.



I printed this one just to see what it looks like. (I've been doing a lot of that this afternoon.) At 9x6" noise is just visible in the smooth metal of the stoplight, but it's less than what I would expect from iso400 film. This was taken at f/2.8, 1/125, and using a half-stop sensitivity boost for an effective iso of 9000.

But there is no perfect camera, and I see no reason why that would change.

[Pavel]

Matthew, I agree with you that highlights and midtones are incredibly clean in your photos. The noise in the sign is not at all troublesome and it may well just be at least in part the texture of the metal, as the noise in the "Do not Enter" sign is far less apparent. This is not the case however in the shadows part of the photo and there both chroma noise and BW noise are quite apparent. You show a scene where the blown highlights are small enough to be perfectly acceptable (to me). However, if the blown highlights were larger or contained critical information, if you used the "expose for highlights" recipe, you would have to underexpose and later pull the some more info from the shadows. And THAT is when you start noticing a problem. Basically, I am saying that you can shoot at high ISO and not adjust the exposure (much) in software OR you can shoot at low ISO and than you have more room to pull details from shadows BUT you can not do both and expect not to get into trouble. D700 is more forgiving than D300. I estimate about 1.5 f-stop. This is a significant bonus that can be enough in many circumstances but ideally I would like 3 f-stops. With 3 f-stops I think I could handle easily and without compromise all the (many) situations where I ran into problems with noise.

Let me be clear. I agree that the way you photograph/process there is no real issue with noise. The way I photograph, the issue comes up quite often. So why do I not just adopt your approach? A good question. It is in part the subject matter, which is often quite different from yours and it is also an issue of how and what story I wish to tell with my photos, which is again often different from your story telling. Both of us challenge our hardware in different ways and so it is only by accident when you and I lust after the same camera.

I will try to adapt the photos to show show the original NEF converted to jpeg, the NEF with adjusted shadow exposure and the final processed photo. But it may be a while.

[Kombisaurus]

Looks like you guys had fun, and also looks like an interesting experiment.

And for the record, if the new megapixel war is going to be fought over ISO, then sign me up for battle! I'd also be happy for Dynamic Range to join the fight too...

[matthew]

I don't know if you've seen it, but Ctien recently wrote an article rather ambiguously titled Expose to the Right is a Bunch of Bull. He says that it's most important to make sure that none of the highlights are ever clipped because the shadows can always be raised later.

[Pavel]

Matthew, that is basically the premise under which I operate, although this approach has obvious limitations (as we just discussed). I wish i could say I came up with the idea, but i have read about "exposing for highlights" several times before i adopted this approach several years ago.

[Zig]

Nice experiment; thank you for letting us all in on it as I find this very engaging.
I reckon it'd be fair to say that in terms of exposure, we all learn to "read" how our sensors behave and expose accordingly, balancing out the noise in the darks and bearing in mind how well our raws are recoverable.
It remains true that highlight detail will mathematically/physically be more "rescuable" and detailed if overexposed, as opposed to shadow detail being so if it is underexposed.
I'd personally opine that allowing one's attention to be too swayed by fears of shadow noise can potentially work against us, whereas a clipped highlight is, if blown, blown for good. "Noise"(and we'd be having this discussion about grain if this forum existed 20 years ago) has so much potential for doing something with: noise on some sensors can look quite film-like...and indeed, there are relatively easy and manifold ways of multiplying darks so that the "shadow-end" of one's zones soften, fuse or block, leaving the mids and highs retaining an effective high dynamic range.
Yes indeed, "exposing for highlights" certainly...remembering that this does not have to be the same thing as metering from the highlights"... Thank goodness perhaps that we do not all have to use slide(tranny) film and be so very conservative with our exposure.
I confess that sometimes, to "make sure" and cover my laziness, I've metered off well-lit grass, set this as my 18% grey point, then added a stop of overexposure. I'll then make a mental note of aperture and speed at say, f5.6, dial off the correction, nip into my default aperture-priority, blatting away whilst comparing my shutter/aperture with the one I made a mental note. I'll that way(though it's a bit rough and ready) get a mental alarm-bell going if there's a danger of underexposure...sounds more long-winded than it is, but just ensures that I err on the side of overexposure.