3D Test ( atmega32 displaying .obj file )

[Dirtgreeb]

I like this. It's a great testament to what an 8-bit AVR with 32K Flash can do. Good job!

[masterpj55]

How did you combine 2 atmega's power?

[zioboby]

25 years ago you could say "OMG AWESOME GRAPHIX!!!"

[Robosos]

its too much for the tiny bloody controller, but nice demo, good job!

[kamilm122]

nieźle ;) robi wrażenie. ATMEGI widziałem w gazecie "elektornika praktyczna"... sa mocne... zamierzam kupić jeden zestawik :)

Jedno pytanko... kodziłeś w C czy Basic? Cy też może gotowe demko z netu :)?

Pzdr dla ciebie ;)

[spryciarze19]

morzesz mi podać strone na której sa te schematy czy all robiłeś sam?

[wrtlpfmpf]

Hmm, now do it without extra RAM. :)

[AndrewFaulds]

maybe the newer atmegas could?

[wrtlpfmpf]

Well those are only availiable in SMT. You can do it without extra RAM by storing the image more efficiently. For example you could store it RLE encoded.

[MiracleKD18]

Is it vector graphics?

[arudson]

Very nice. Are you using fixed point math? What kind of SRAM chips are you using?

[krzysiek007]

Unfortunately it's not fixed point math - currently it's exceeding my knowledge (maybe in future :)

I think with fixed point, it will be working faster :D

As SRAM's are used Dallas DS1249Y - I know they are a little too big, and they are NV but I had them unused in a cabinet.

[AmstradExin]

Yeah, really looks slow compared to what an Atmega32 can do. Especially when you use one only for graphics. :D :D

[mast3rbug]

I have done a 3d routine with fixed point math. No division, no SIN, no COS. All in assembly language with a single CPU. It's really fast. My engine is compatible with 3d Studio Max.

I don't have video of my output on a screen but I have the example on my RGB virtual clock. Do a search in youtube for "rgb virtual clock final version" This will be the first link. The real time 3d process is at the end of the video. Feel free to give me comment about that. Anyway, nice work!

[Amund7]

@krzysiek007 Fixed point is easy: To store: fixedN= floatN*(1 shl x) and to read: fixedN shr x (where x is the number of bits behind the comma). The whole point is to do all the math with the full numbers until you plot the pixel, only then you shr. If you use matrix math for rotations and projection this should be many times faster, unless this chip has a really fast floating point processor, which I doubt...