I think your problem is more with the size of your redo log file than the buffer. Let me explain.

2 files of 200 M each. When one gets filed up, a logswitch/checkpoint occurs which cause DBWR to move all dirty buffers to the datafiles. In the meanwhile your other log file is getting filled up fast enough and it becomes full before the checkpoint on the first file is over. Naturally, when the second file fills up it has to wait to clear out and causes the log_buffer to fill up which in turn causes users to wait for space in the log buffer.

Check your alert log file for "Checkpoint not complete" messages. If they exist then reduce the size of your log files/ increase number and also analyze your LOG_CHECPOINT_INTERVAL and LOG_CHECKPOINT_TIMEOUT parameters.

Nizar.