Hacker News Reader

Comment by pedrocr | Hacker News Reader

People used to make similar arguments about 35mm film. Now that that's firmly surpassed I guess 4x5 will do. 4x5 is only 16x the area of a 35mm sensor so it's not too far away, you will have to soon move on to 8x10 to make the same argument :)

View on HN

Current APS sensor technology is around 15 stops of dynamic range and 50MP. 4x5 sheets of Portra can resolve 100-300MP with 18 stops of dynamic range.

Digital has beaten 135 film, but it still hasn't beaten 120 / medium format in several key areas. I would be surprised if digital beat 4x5 in the next 10 years because there's no substantial market for such a product.

You're just making my point exactly. You're getting 6 times the resolution out of 32 times the area. 35mm sensors are already common and that's already twice the area with better SNR. Maybe not all the 18 stops but close and it's also all semiconductor technology which improves for many other purposes than making sensors. There are also computational techniques to make the equivalent of a larger sensor from many small ones. We're coming at this from too many directions to not solve it. We'll soon need to have this discussion with 8x10.

The resolution of the sensor is pretty irrelevant. Camera and subject motion, the MTF of the lens, and diffraction are much bigger factors (once you get to a certain point). You're almost diffraction limited on modern APS-C cameras at f/8! The motion of your hand and body will cause a point of light to illuminate two sensor elements even at 1/8000 shutter speed.

Basically, DSLRs require perfect conditions in the field to actually achieve the sharpness stated by the camera manufacturer. A 4x5 sensor allows much more slop.

So you're doing 4x5 handheld? I don't understand your point. If there's anything digital clearly wins at is at getting you better results than film for a given physical constraint. It has better sensitivity so takes better advantage of the light, it has more density per unit of area so gets you more resolution per unit of camera volume or weight, etc. The only thing film still wins at is being able to cheaply build larger "sensors" since film doesn't get more expensive to manufacture larger. And as sensors have gotten better it's gotten harder and harder to actually realize a benefit from larger film up to a point where you need a large cumbersome camera in a very sturdy mount and really good technique and process to have a shot at it. I expect that within the next 5 years multi-sensor/multi-lens cameras and a bit of software to finish off the last few advantages there still are[1].

[1] https://gearjunkie.com/multi-lens-light-l16-camera

"People used to make similar arguments about 35mm film. Now that that's firmly surpassed"

Nanoparticle films are coming out. Gigapixel resolution in 35mm format.

Meanwhile semiconductors are already so good that we could build that today already. Nokia's 2012 PureView sensor's density was over 400MP in 35mm and that's already 4 years old. I'd bet on the digital side for this.

Eh, we can get into the picometers with analog molecular films. Got any semiconductors at that scale?

Are there any actual films that can outresolve an atom? That would be quite cool.

No, but we've got films in development where the individual grain size is just a few atoms. Got any semiconductor display techs capable of that small size?

There's nothing fundamental stopping semiconductor technology to reach those feature sizes and a fair comparison is between technologies at the same development stage. If you use the proxy of what is commercially available today CMOS sensors beat film. There are still reasons to use very large film, mostly because no one develops large enough sensors (and even if they existed cost would be very high), but the advantages are being overstated.

"There's nothing fundamental stopping semiconductor technology to reach those feature sizes"

The laws of physics most certainly are THE fundamental thing in the way. Electron leakage, current ripples at nanoscale, capacitive interference. There's a lot more involved when it comes to semiconductor technology.

Reply on news.ycombinator.com