Hi, I'm trying to wrap my head around how to implement a depth/fake 3d system in a 3/4 top down style game.

I started with looking at one of the most obvious examples: A Link to the Past.




This seems to work via a pretty simple optical illusion and some basic one-way collision detection in regards to falling off cliffs, no big deal there.

Then I looked at the PS1 game Alundra. This game is in (more or less) the same perspective as A Link to the Past, but with one big game changer: the ability to jump.
In the world, you can jump on any piece of scenery you can reach:



You need to press forward when you reach the table's height in order to "catch" the table, if that wasn't obvious.


The sprite moves on the y-axis when you jump, but it's different than "normal" movement. For example, you can stand as close to a door as you can without going in and jump. You won't go into the door, but your sprite will move along the y-axis past the door trigger (as well as be rendered above the top of the door frame) but you will not enter the door. If you move your character forward a single unit with the d-pad (or whatever) you will of course enter the door. The game also supports horizontal slopes, but I think they are locked into a single angle, so probably not a huge mystery to solve there.

The result of all this is a really really nice 3d effect.
So naturally, my question is: What sort of options are there for this kind of "robust" faux-3D? With no real z-axis, it seems really tricky!

I'm generally just looking for a discussion on the theory behind doing 3D like this and maybe some pseudo-code, but if it matters I'll probably try to implement it with C# in Unity (forgoing their collision and physics systems more than likely).

Note that you can have 3d gameplay with 2d graphics and vice versa.

In A link to the past, there's just a flat timemap (judging by that screenshot as I haven't really played it, at least not for many years). In that case, walls are just blocked tiles. In alundra, the tiles on the map have a height. This means each tile has a floor and wall texture/image. When you draw, you draw back-to-front. For each map 'row', you draw tiles first, then objects. If an object spans several rows, it's drawn after the last (lowest on screen) row. When you draw a tile, you must first use its height to find the y-position on the screen. If its height is lower than that of the tile in the last row, you fill in the distance using the wall texture/image.

Finding the y-position on the screen is just map_y + height.

Alright, I think I have a pretty decent idea of how to get something similar to Alundra going now.
I had thought about having a "fake" z-axis before, but I figured I'd run it by some more knowledgeable people before proceeding.

I'll update this thread if I run into anything interesting/frustrating.
Thanks a bunch guys!