Eh? The quads don't get scaled at all - the main mesh is just positioned far enough away from the camera so that it's all shown on screen. I create a quad depending on the image in question. The bigger the image, the bigger the quad. Is that wrong? That's how I assumed single surface systems work.