Thank you, it works. It doubles the avaiale memory form 160MB to 320MB. This is enough for 20 track with 16MB each. For audios with 22kHz this means 5:30min tracks (OK!), for 48kHz this means 2:45min, which is still a little bit to less.
The app already exists on Windows System. And here I had to care less about memory consumption. All Tracks were saved in INT-Arrays, where I could manipulate each sample-value easily. Even a second INT-array of same size was possible to keep track related interim results. This made calculating the sum signal extremly fast, when only one track was manipulated by the user. Memory consumption here: 2 INT-arrays(10.000.000,20) = 800MB
Now I see, it is necessary to build a more complex data-field and this causes a lot of more calculation in case of user changes. I think performance will get my next problem.
thanks again for the code line.