time constraints

[tiziano]

I know that it is required to stay within the timing constraints imposed by the pipeline, but what happens if, for any reason, they are not met?
I simply loose a data block? or something worse could happen?

Thanks a lot,
Regards

Tiziano

[gschalk]

Tiziano,

This is a good question. If your system consistently uses more time than one sample block, then it is simply too slow to run that BCI paradigm. In this case, you need to reduce the computational complexity of the BCI system, or need to increase the speed of the computer.

If your system is using more time than one sample block only occasionally, then the system does what it otherwise would. Once the data are processed, the next block of data is fetched from the data acquisition device. It then depends on the data acquisition device and the implementation in the source module what it means that you are now picking up data somewhat late. In the g.USBamp implementation for example, there are buffers so that data are not lost. The Neuroscan implementation does not have buffers, but BCI2000 gets data from Neuroscan via TCP/IP, and TCP/IP has its own buffers. Thus, for these two and other examples, no data will be lost (at least assuming you pick up the data in a reasonably short time), but for other data acquisition devices, there may be data loss.

Gerv