P.S. I should add that the smaller pixel size on the camera isn’t an insurmountable issue. Like most modern SLR’s the noise at the lowest ISO setting is perfectly acceptable, and you can hardly tell any difference between cameras with big pixels and small pixels. They do have worse signal-to-noise, but if you aren’t whacking up the gain, you can hardly tell.

The problem is that small pixels give you less leeway. If you are shooting in lower light and therefore having to turn the gain up in the camera (by turning from say ISO 100 to ISO 400+) you start to notice the effects of the noise at a lower ISO setting from cameras with smaller pixels. So where a full-frame camera with large pixels can probably cope with shooting at ISO 400 and still give the silky smooth “gloss” look we usually go for on RE, a camera with smaller pixels might already be showing visible amounts of noise.

If the in-camera sharpening is turned on, that can further sharpen and exacerbate the noise. The in-camera sharpening is tuned to sharpen the finest details that the sensor can resolve, because it is set up for producing 6 or 10 or 15 megapixel output and they don’t want to obliterate fine detail by sharpening over a two or three pixel radius. However, as we’re producing web-sized output at more like 1 megapixel, this isn’t the optimum choice for us. It is better for us to sharpen at a scale of three pixel groups (on the original 6 megapixel output) as that’ll be better matched to the final 1 megapixel output. Sharpening at 1 pixel scale just emphasises all the noise and prevents the usual smoothing effect of a resampling down to 1 megapixel from having such a nice effect.

Various cameras also have noise reduction algorithms which they do on readout or when converting RAW to JPEG in post production. This rarely performs as well as doing an unsharp mask at the right radius in post production, and resampling. Again this is because it is tuned to produce detailed output at 6 /10/15 whatever megapixels instead of at the web-sized final output 1 megapixel.

So these cameras with smaller pixels can absolutely produce pictures in the “glossy” RE style, but you have to know how to set things up to allow it. The most important factor is the signal to noise in the very first step, which means turning the lights up so the camera is functioning at ISO 100 with no additional gain needed. (In fact we tend to light to about f/8 at ISO 100 because the lenses tend to perform best at about that aperture, which is unrelated to the noise effects but does influence the crispness of the final image). The second factor is to turn sharpening in-camera or in RAW-to-JPEG steps right down because the algorithms used do not know that we’re aiming to output at a reduced size of around 1 megapixel.

Ironically, the march to ever greater megapixel sizes on the image sensor makes it harder and harder to take these smooth images, especially at the cheaper end of the market. I have a seven megapixel Casio snapshot camera whose noise performance is so bad that the images have to be resampled down to 3 megapixels to look decent even when the images were shot in the full glare of the mid-day sun in the desert in Jordan at ISO 100.

Thus far the big SLR manufacturers have done pretty well in keeping the noise down as the the pixel sizes have decreased, by various tricks and improvements in manufacturing techniques to reduce the readout noise of the chips. There is only so far one can go down that line before physics will bite your backside and start compromising the noise, though, so I’m in no hurry to change the studio 5D for a 5D mark II. I’m just not convinced there is anything to be gained from going to a 21 megapixel sensor when your output is only 1 megapixel. The key selling point of that camera to me is the auto sensor clean function because the 5D’s sensor attracts shit like nobody’s business.

Cheers, Hywel.