diff --git a/decoding.txt b/decoding.txt
new file mode 100644
index 0000000..ec7516f
--- /dev/null
+++ b/decoding.txt
@@ -0,0 +1,29 @@
+Decoding:
+
+4 wpm to 50 wpm adjustable filter. 
+
+Record input, place into ram then process strips of it? Establish WPM and then determine what it says.
+
+Need to establish noise level. Measure for a bit, then set variable as an average. Start measuring when goes above noise level for x samples?
+Need to ensure this is added to start of first character so that issues not seen.
+
+to record, need to establish length of signal and an appropriate sample rate. IF input == noise level for longer than x then stop recording. 
+
+As adding into ram, could need to strip out 'junk data' so can reduce memory load. Need to ensure a low is kept between high and lows as well as between words.
+Can we do something to affect the number of highs? 
+
+Does the WPM need to be identified?
+If using a constant sample frequency, can try to compare dits to dahs. 
+Identify HIGH, record number of samples before a noise level is found
+Compare to previous HIGH length.
+If one longer than the other by x% then dah found. 
+Pass through recording identifying and then converting samples to dit and dah.
+Try to work out pattern for message. 
+
+WPM might be required for message establishment? Could be determined through length of 0!!! Hmmmm avoids calculating WPM, increases RAM.
+
+Concern - how quick is this process? Can it cope within 2s? Probably..... 
+
+Once words identified, form a message to send to screen. LED changes colour depending on operating mode?
+
+Error handling???? When to reject a series of characters as noise? - might need an option to rebaseline the noise level.