Oversampling

Geoffrey_Woodget

I realize this isn't a scolastics site, but can someone explain to me, in simpleton language, how to do the math here?

So I'm watching Daz's short video on "how To Render Faster" https://www.youtube.com/watch?v=t1UWpBcPRs0

At 2:56 min in it say's to change the 16x9 aspect ratio -1920 x 1280 by 50% larger so that after rendered you can scale it down by 50%. So, he say's 1920 becomes 2730. Huh?  How do you arrive at that? (map reading is where I excelled on the Iowa Test Scores)

Thank you

y3kman

1920 x 1080 = 2,073,600 pixels

150% of that contains 3,110,400 pixels

To keep the 16:9 aspect ratio, you'll need around 2352 x 1323 to get that number of pixels

The guy in the tutorial probably just made a guess/approximation.

Tobor

The video uses the wrong math, and should probably be redone.

When entering the new pixel size, simply do the math inside D|S: Type 1920*1.5, and D|S will do the multiplication for you: 2880. If you have the option set to constrain proportions, the height is automatically set for you.

You don't want to multiply the area of the pixels by 1.5, as that may not produce enough of an oversampling. Multiplying the X/Y values by 1.5 is not the same as multiplying the area of X/Y by 1.5:

Original X:100, Y:100 pixels = 10,000
150% oversample X and Y: 150x150 = 22,500
150% oversample area of X*Y: 15,000

The exact amount of oversampling is not critical, but you want it to be enough so that when the image is down-rezed in your graphics program the unconverged pixels will be filtered out by the reduction sampling algorithm. You should experiment with ideal sampling sizes and reduction settings before doing an entire set of renders.