32 bit float in Fusion

I'm trying to use 32 bit float maths in Fusion.

I noticed that the only place where i can configure it is in the "Fusion" menu, under frame format. There is no such setting in the MediaIn node (in Fusion standalone there is the option in the loader node).
But when setting to 32 bit float it still looks like it's working in 8 or 16 bit integer, no negative values. The waveform shows a signal cropped at 0.

What i'm missing?

Thanks!

Or, maybe i’m using the wrong node.
Is it supposed the merge node in difference mode to crop at zero value or to go negative?

Difference is, as name suggests, difference between two values. Calculated as abs(A-B), so it can't be negative. Use minus if you want negatives also.

I can't seem to get the color depth options in the Settings menu to do anything at all. Fusion always reports 32-bit color unless I use a Change Depth node to set it to something else.

To perform a Subtract, use a ChannelBooleans node.

Got it.
Thanks Bryan and Hendrik!