[robski]

Annette it looks like you have got used to the improved noise performance of the 350D and have forgotten how much noiser the 300D was.

From Bob Atkin site - As I am from an electronics background I would of said exactly the same.

"You can't make a noiseless sensor and the smaller the pixels, the higher the intrinsic noise. There's a lower limit to noise which is determined by fundamental physics and includes things like photon shot noise which is determined by the statistics of photons hitting a pixel, plus various forms of thermally induced noise. At some point you'll hit a limit to the number of usable pixels on an APS-C sized chip (24mm x 15mm). It's probably in the 10-12MP range. People like more pixels and will usually pick a 12MP camera over a 10MP camera over an 8MP camera. At some point the only way you can up the pixel count without lowering image quality (by increasing image noise) is to go to a bigger sensor. That's why 8MP DSLRs outperform 8MP consumer digicams with 5mm x 7mm sensors. Size matters."

By understanding the physics engineers can reduce the noise but will never be totally rid of it. Another technique to improve the signal to noise ratio is to design the light sensitive part of the sensor pixel to be nearer to the surface of the CMOS so that less light is lost in reaching it. Again there will a limit to what they can do.

If you avoid under exposing this should help keep noise levels low. You will notice with consumer digicams which have a small sensor the ISO range is limited for the very reason to keep noise levels acceptable.

I will often use a noise reduction program - in my case neat image.
firstly to reduce any apparent noise in the image. I either use the auto method or apply a camera noise profile that matches the ISO setting. I may also tweak the default setting such as sharpness and noise in the luminance channel. The other reason will be purely to get the file size down for up loading to the web.

Noise and image can only be 100% separated if you know exactly what one of the components is. As the program has no idea what your image looks like it can only work on the characteristics of camera noise. Noise by definition is random. Techniques can be used to work on truely random noise. Unfortunalely camera noise is not truely random. The noise tends to fall into defined patterns dependent on the source of the noise. Thermal noise (temperature of sensor and electronics) - hot days will give you more noise than cold days. Even Gamma rays from outer space can create noise. The sensor amplifiers can produce their own noise plus amplify the sensor noise. The camera noise profiles have models (information) of strenght and shape of these patterns so that the program can try and find a match and then reduce or remove them.