A new release is coming in the next few days. I've fixed a few memory issues, which should lead to a lower memory consumption. Nevertheless is the memory consumption for large mdf files inacceptable if you read all samples of one channel. The problem here lies in the automatic conversion of the values which I scale back to their double-precision using the associated conversion formula.
My next step will be the implementation of a sample reduction algorithm, which should improve the memory usage for the visualization of channels. For faster/better signal handling the imoplementation of memory windows, which will allow you to push a window of a number of samples through the mdf file.
Last but not least I implemented a small signalviewer for MDF files as a proof of concept. I will post this here later. So stay tuned for my next post.