Hacker News Reader

Comment by reneherse | Hacker News Reader

Lower noise levels at high ISO are an advantage of FF. Very useful for shooting action in low light (e.g. sports and dance), where you need high shutter speeds to freeze motion.

View on HN

> Lower noise levels at high ISO are an advantage of FF.

You are correct, except I don't like the way you said it. I think it is more accurate to say that full frame has lower noise levels because they tend to be able to achieve the equivalent of lower ISO.

In order to get equivalent ISO, you need to multiply it by the crop factor squared. (Squared because the sensor area is two dimensions.) So for MFT 2x crop factor, that means 4x. Which means a MFT sensor at 800 ISO has the same theoretical noise level across the frame as a FF sensor at 200 ISO.

This is another area where the way we describe crop factor does a huge disservice to practical application. By pretending that smaller sensors are just cropped versions of larger sensors, we have to do a pile of stupid math in order to find equivalence. And it causes no end of pointless internet arguments.

https://petapixel.com/2017/03/29/dispelling-myths-around-dep...

I think what you meant is that low noise and high ISO are advantages of large pixels. Larger sensors just happen to have more such pixels. If you where willing to eat the reduction in resolution, you'd get the same noise and sensitivity from any size sensor. For example, a 10MP APS-C sensor has the same sensitivity as a 25MP full frame sensor, all other things held equal.

Per pixel. Not overall.

25MP (or any resolution) full frame sensor with a (say) 1.8 lens will gather more photos than 10MP (or any resolution) APC sensor with a 1.8 lens.

The amount of light falling on each sensor pixel is the only thing that matters in digital photography. The roof of my house also gathers a lot of light but it's not a good camera.

The quantity of photons captured and usefully turned into data by the total area of the sensor is what matters in digital photography. The physical size of each sensor pixel is irrelevant, except insofar as current technologies favour physically larger pixels in order to capture larger quantities of photons before the pixel is fully saturated.

Yes, that’s a more precise description. For a given resolution and ISO, a larger sensor has larger pixels, which generally produce less noise than smaller ones.

But it's not actually the physical size of the pixels that ultimately matters, it's the total amount of light captured by the sensor.

Thinking about the individual pixels of a sensor is tempting but causes an unfortunate chain of erroneous assumptions.

It is true that (a) larger sensors have had an historical efficiency advantage, but it's progressively diminishing, and (b) larger sensors tend to be capable of lower area-equivalent ISOs. If you could build an MFT sensor capable of ISO 25, and used comparable glass, the full frame advantage would be eliminated.

Per pixel noise should never be what matters. What matters is how much overall noise is in the image. To put it another way: how noisy are these sensors after you scale their images down to 8 megapixels?

Large format prints aren't scaled to 8MP.

You missed the point. That was just an example to aid the explanation.

Noise only matters in the context of output. If you’re printing—again just an example—a full magazine spread, the noise level in the final result isn’t fundamentally affected by whether the sensor was 30 or 50 megapixels; FF or MFT. Either way it is ultimately scaled to the print size and print resolution. Nobody except the photographer cares what the noise looked like at 100% pixel zoom. All that matters is the relative appearance of noise in the final product.

Reply on news.ycombinator.com