[sdiy] MIDI State Machine diagram - please comment

[johnspeth at yahoo.com]

> > http://mysite.verizon.net/jspeth/MIDI_FSM.pdf

Neil's list is a good summary of things that are wrong or not quite right.

> - In your "status unknown" state you don't handle System Realtime messages, 
> while you do in other states.

You're right.  I'll correct that.

> - Why treat Note On and Note Off differently?  They have the same structure, 
> just different status byte.  As Achim suggests treat them the same, and then 
> interpret the collection of three bytes later on - its also much easier to pick 
> up on the 0-velocity note-off command.

This issue is a prickly subject.  Let's not confuse behavior with implementation.  I'm attempting to capture behavior.  My ultimate goal is to go from a behavioral diagram to an implementation with little extra work.  I'd love to know if anyone knows of a tool or language that can take state diagrams as input and output C or C++ code.  IAR's VisualState is a pricey tool for that purpose.  I'd prefer a free one.  Anyone?

Nonetheless, I readily admit that behavior can be captured using an infinite number of expressions of state machines.  The NOTE-ON/NOTE-OFF similarity is an obvious way to simplify my state machine.

> - In the Sysex state what do you do if you receive a status byte other than 
> EOX?  The official line is to abort the sysex.  Specific to your diagram, what 
> if a Note On status byte terminates a sysex?


You're right again.  I'll correct that.

> - What about all the other MIDI messages?  There is lots of useful information 
> coming across the MIDI cable.  Seems odd to support Sysex but ignore pitchbend 
> or even mod.wheel.

Yep.  My intent was to build a custom MIDI interface for my Prophet.  The digital interface doesn't handle wheel inputs.  There's a separate analog interface that I decided not to use.  I suppose my FSM should have included the required MIDI Implementation Chart.

> - No mention of channel filtering for the note messages.  Are you supporting 
> OMNI ON mode only?

You're right.  Channel filtering is implied.  In my device, the channel will be fixed (no front panel).  I can add a note about that.

> Look forward to seeing the update!

Rev 2 is at http://mysite.verizon.net/jspeth/MIDI_FSM.pdf.

Thanks for all the comments.

John Speth.