Building a Digital Organ using a few years of hard knocks

 For a lot of years now I've been developing electronics, techniques, and designs to optimize designs for digital organs.  With my partner Jon Orwig of Evensong Music, we have developed some amazing sample sets of American instruments.  We're super happy with this part of our business.

I've been developing electronics to create MIDI interfaces for digital organs for the last several years.  The designs have had many iterations, but I'm narrowing down to a pretty optimized design.

 This review is for a practice/teaching instrument for St. John's Lutheran church in Orange, CA.  The plan is for a 3 and pedal manual instrument using our St. Olaf, Minneapolis sample set.  One challenge for this instrument is that it's going in the music office, which is a house built in the 1920's, and has very narrow doorways.  A traditional case, even ones that are designed to go through 32" doors, will not work for this application.  A "flat pack" design has to be done to allow it to fit into the space.

The elements in any organ build are

    Key Action (Keyboards and Pedal)

    Stop Action

    Combination Action and Control

    Shoes

    Physical mounting and appearance case

    Sound generators/relay (pipes or Hauptwerk.  This is another discussion altogether)

 Key Action

 A critical and one of the most costly elements of any digital organ build is finding the keyboards.  I'm lucky enough to have a great relationship with our local builder Ryan Ballantyne.  He built the V/150 instrument in the sanctuary at St. John's, and through his business has access to lots of parts.  He has given me a 3 manual stack of Peterson keyboards for this project.  The keyboards have a full complement of pistons as well, already mounted on strips, so this makes things much easier.

The Peterson key switches are arranged into octaves.  Traditional setup is to hook the commons of the keyboards together.  However with this structure, it makes it easy to do a matrix scan.  

One of the things that I found when I first looked at traditional organ design is all the time and energy that was spend in making wiring harnesses.  Using my background in consumer and commercial electronics design, I set about to optimize these designs so that that time didn't need to be spent.

This matrix scanner (yes, I know the diodes are in backwards - the parts I used were able to be rotated 180 to point them in the right direction) allows the keyboards to be set up into 5 groups of 12 plus the 61st key as a separate input.  

 

For each octave, only the 3.96mm connector, the column connection, and the ribbon cable connector need to be installed.  At the end of the chain, a fully populated board with the microprocessor are installed to do the scanning of the entire keyboard.  This removes the need for any harness other than a 16 pin ribbon cable.  The resulting setup looks like this

 

 This allows the scanning of a single keyboard, but doesn't turn it into MIDI messages yet.  The system I've designed has a couple of busses which are used to turn bits scanned into MIDI note on and off messages.  The 10 pin interface used for this part of the system is a multi-drop interface which will support 256 slots (up to 4 61 note keyboards).  An Arduino processor interfaces to two of these 10 pin busses, allowing up to 512 separate MIDI events in a system.

There are also in this design two separate interfaces for a 74HC597 style serial shift register.  This is not used in the keyboard part of this design.

The pistons in this design are wired into a transfer board which turns individual wires into a Ribbon cable connection, which are then scanned just like another keyboard

Pedalboard

Pedalboards are more difficult to source, and are more expensive and cumbersome.  A place to get good pedalboards is from discarded instruments.  In particular, if you can find them, Rodgers and Allen instruments from before 1965 or so were built just like pipe organs, and use high quality pipe organ components. This includes keyboards and pedalboards.  For this project, my friend Ed Ballantyne found an old Allen from the 1950s available locally.  (These can sometimes be found on eBay, but as with everything, usually shipping is the issue.  Usually they are 2 manual instruments, which are fine for student practice, but my opinion is that 3 manuals is the minimum necessary for serious study, and adding a 3rd manual to an existing case in my experience has usually been a hack)  I have appropriated the pedalboard for this project.  Incidentally, those old Allens are beautifully built, and it was a shame to pull it apart.  Those guys built a functional organ emulator without using a single transistor or active component.  The entire thing was wire, magnets, and Resistor/Inductor/Capacitor.  Nothing else.  As an engineer, I find it amazing.

For this pedalboard, I used a common return for all notes, as I had developed a pedal scanner (along with the matrix scanner discussed above) for a project with Bob Knight of Knight Organ Company in San Diego.  While the pedalboard could easily be folded into the common processor with the keyboard, it's more convenient and robust to have it connect to the system through its own USB cable rather than having a ribbon cable that could be incorrectly installed.  This also allows the user to move the pedalboard more easily. (Full schematic can be accessed here)

This PCB and software design also includes an interface for shoes and up to 8 toe studs, but the shoes in this design will be handled by the keyboard processor, as the shoes are physically attached to the main console rather than being attached to the pedalboard.

 

Shoes

 Shoes are another thing that is pretty difficult to procure.  I had purchased a pair of Allen shoes many years ago without a home.  They now have one.

Allen shoes are optical.  An old school incandescent light is on one side, and a light sensitive resistor is on the other side.  A blade is operated by the shoe and varies the light coming to the photocell. Due to how Allen did their design, They wanted a non-linear response.  

The response was a bit too nonlinear to make sense of with a simple 10 bit ADC, so I cut off an edge and taped off part of the other to make a more linear looking opening.

Even with that, the response of the sensor is still very nonlinear, but it is in a range that can be read by a 10 bit ADC.  I'm using a look up table to return more a more linear profile.

Case

I've been working for a while on how to build a cost effective table.  I have previously made a table and a bench from a single sheet of plywood, but for this project I wanted something a little more robust, including a full panel above the shoes.  The design still fits within a single sheet of plywood, but there isn't enough left over for a bench.  This is the cut plan

 And the general side view layout

Granted, I used some scrap I had hanging around, but this table cost less than $150 in lumber and hardware to build.  I'll need to apply some finish.

 

Stop action is in process, more on that later.