I see Blender has a multitude of file formats for rendering. I want to render an image sequence with the best quality and the least amount of memory so which file format would you render still to and why?
BMP, DDS, Iris, PNG, Targa, Targa Raw, Cineon Etc.
I tend to use .png
Thanks Richard. Me too. 
I love .png format but was just interested to see if there was a better format that I maybe should be looking at. Anyone else use something different?
PNG. Lossless says it all. Smaller than tga without compromise.
I use TGA for anything I do professionally, PNG for my own personal stuff.
Why TGA for pro work?
Just an old habit from work.
Slightly off topic…
png allows for compression level – at what point does it become lossy?
in other words – why prompt for compression level if it makes no difference on quality? Just compress as much as possible…?
From what I understand if you compress a .png it will lose some quality, due to the removing of some of the pixel information, but lose no more quality beyond what you compress it too, in other words if you compress it it will remove some of the pixel information but subsequent uses of the .png files will not lose any quality (lossless) whereas .jpg’s which are (lossy) will continues to lose quality with use. That’s at least how I understand it.
OpenEXR is the highest quality option, as far as I can tell. If you have more work to do and your subsequent tools support it, I’d say it is the way to go.
-rking
At none. The compression depth selection is soley a speed issue. Try to save a 3000x3000px png in Blender with uncompressed png and with highest compressed png. It’s a save difference from instant to up to 15-30s.
In short, no. In long… nooooooo. 
Bollocks. All OpenEXR offers over PNG is to store layers. So it might have more features, but not the highest quality.
Again, PNG is a lossless format. Nothing is lost.
PNG uses DEFLATE, a compression also used in zlib, which is a mix of LZ77 and Huffmans algorithm. A similar method is used in ZIP.
Wouldn’t be great if you compress a file with ZIP and loose information huh?
It may not be higher quality, but surely since .exr has a much wider range than .png it can store more information, and is therefore useful if you intend post processing rendered sequences?
Better quality = bit depth.
If you are rendering a sequence that might be re-used as a texture in another render system I would suggest rendering to a TIFF sequence. It is lossless, has an alpha and can support layers like OpenEXR. But the main reason to use TIFF is if you plan on working in the Renderman realm. TIFF seems to be a dominant file format for 3Delight and some other Renderman compliant render systems. They can’t/don’t read PNG files.
Is support for 16bit / channel color needed? Is Blender 8bit / channel? To my knowledge PNG and JPG are 8bit TIFF 16bit.
True, OpenEXR nateively supports floating point, while PNG only supports integer.
For the sake of completeness though there are PNG extensions for that.
Since the PNG 1.1 Extension specification it simply stores floating point values as strings. “sign”+“fraction part”+“.”+“exponent part”
If it’s practical though…
I use .png; can’t actually tell the difference between it and openExr.
Is support for 16bit / channel color needed?
Yes, otherwise you can get visible banding within gradients.
Your monitor can’t display the difference. The benefit is in manipulating them.
This has actually been an informative discussion. Thanks everyone.