Solutions for a very high number of inputs

It's not so hard,

I'll assume that 1mS latency is good  enough, so you need  a 300kHz clock on your shift register, and thats assuming that you use just one serial stream for all your inputs.

A bigger problem will be wiring it all up, if I were doing this I would design  a little board with about 32 inputs (using  a 64 pin micro) and put these where physically convenient to reduce the amount of wire. Then link the little boards to a cenrtal controller using CAN, RS485, or whatever you prefer.

An STM32F100R4T6B is £1.98 from farnell, has 64 pins so will give you 32 inputs and a serial link with no need for shift registers. You'll need to design a board but its not too hard.

Let me know if you fancy going this way and I might be able to help a bit more.

MK

I actually designed something similar to this several decades ago.  First, some general comments:

1.  Don't worry about "near real-time response".  MIDI only runs at 31.25 kb/s, so you're going to have some delays unless you use time-stamping.  A small amount of delay is actually good because real musicians don't play notes at exactly the same time and if you synthesizer does it will sound artificial.  The same goes for pitch -- I've heard synthesized music sounds more natural if instruments are slightly out of tune with respect to each other.  Of course, you don't want to go too far: "What's the difference between the first chair viola and the last chair viola?  Answer: A measure and a semi-tone."

2.  Do worry about switch bounce.  Mechanical switches do not produce clean on/off transistions.  They tend to bounce between on and off until the contacts settle.  You'll need to filter out these transistions.

So let me tell you what I did back then.  Since the technology was rather different, I'm going to borrow from The Four Yorkshiremen.  I wanted to build a simple polyphonic electronic organ using two 37-key surplus keyboards, with a one-octave overlap to give the usual 61-key organ keyboard range.  These keyboards didn't have switches: "Your keyboard has actual switches?  Pffaaa -- I had to make my own switches out of paper clips with wire insulation stuck over them, and glue each one to the end of a key -- I even had to make my own glue."  (OK, I made up the part about making my own glue.)

Each key had a wire-wrap wire soldered to it and the wires were connected to the inputs of five 74150 16-to-1 multiplexers.  "Pffaaa -- back then we didn't have CMOS or even LS.  In fact, Schottky hadn't even been born yet."  Basically, the five 74150's plus a 74LS151 acted as an 80-to-1 multiplexer to select a key.  An advantage of TTL is that you don't need pull-up resistors on inputs: they automatically float high if not connected.

Now the interesting part: I needed to debounce the inputs but I didn't want to add 80 resistors and capacitors.  Besides, debouncing with RC networks requires that the digital inputs have hystersis (Schmitt triggers) which 74150s don't have.  So I did the filtering digitally.  I used a 256x4 bit SRAM as a bank of 4-bit counters and then time-multiplexed logic to update the counters so that each counter was visited every 250 us or so.  Each visit would read a key's counter value into an up/down counter, increment or decrement the value depending whether the key was pressed or not (saturating at 0000 or 1111), and then write the updated value back into RAM.  If the MSb changed, it meant that the key had been pressed or released long enough for the bouncing to settle.  When MSb changed, the board would interrupt my computer (a Heathkit H-8) which would update my synthesizer board to add or delete a note.  The keyboard scanner was a couple dozen TTL and LSTTL chips and ran at 2 MHz -- a nice solution for 1977 wire-wrap technology.

A scheme similar to this might work for you.  This being the 21st Century, I'd get a microcontroller board with lots of I/O pins such as an ST Discovery board and then add external multiplexers to get your 300 inputs down to 16 or 32, whichever works for your board.  You can do the filtering for 16 or 32 inputs in parallel.  If you don't know the trick, let me know.

This seems like an attractive solution from both a wiring and programming aspect. With a microcontroller for each small area, I could make each of them responsible for figuring out their own state changes, then send only actions which need to be taken on to a central MCU for conversion to MIDI. Correct?

I must admit, though, the thought of having to design a board is a bit intimidating, as is manually soldering 10 64-pin chips to that board.

Can you give schematic / block diagram? to understand better , for me it is difficult to understand high profile terms.

Regards ,

Sanjaylmbr@gmail.com

Sanjay Limbore wrote:

 

Can you give schematic / block diagram? to understand better , for me it is difficult to understand high profile terms.

Regards ,

Sanjaylmbr@gmail.com

Thank you for your interest.  I probably still have the schematics and a write-up somewhere in my archives.  However, even if I can find it, it's hard copy (predates word processors and schematic editors) so I'd have to get it scanned into a PDF.  Plus, when I see the write-up I may be too embarrassed to share it -- this was some 35 years ago.  If this is important, I'll go to the trouble.  Please let me know.

I think my design was a good 1978 solution for 80 keys.  However, I doubt you'd be able to get some of the parts these days.  Today I'd go with one or more micro-controllers and debounce in software.

John...those were the days when ingenuity was really tested, bought back memories...Rod

Hi,

Single line diagram could solve my purpose, it it is difficult / trouble to find old docs , I want to built low cost organ for myself. since in my country it is difficult to get component which are normally not used . whatever easy do it for me.

Regards

Sanjay Limbore

I have never faced a problem like this but I would like to propose a different approach :

Multiplexing.

It's how the PC handles its keyboard.

Each organ normal keyboard has 88 keys that could be handled by a 11 bits of input and 8 bits outputs (19 lines).

The full organ could be handles by a 18 x 18 matrix (324 keys).

One of the advantages of this approach is that the cabling would be done in the keyboard itself with fewer lines going to the processor(s).

Sorry, my last message should have been addressed to you, Alex.

Aryldo Russo wrote:

 

I have never faced a problem like this but I would like to propose a different approach :

Multiplexing.

It's how the PC handles its keyboard.

Each organ normal keyboard has 88 keys that could be handled by a 11 bits of input and 8 bits outputs (19 lines).

The full organ could be handles by a 18 x 18 matrix (324 keys).

One of the advantages of this approach is that the cabling would be done in the keyboard itself with fewer lines going to the processor(s).

That's a good solution if you have isolated switches.  In my 1977 design the switches had a common ground, so multiplexing had to be done using ICs.

For a polyphonic keyboard each switch needs a diode.  http://en.wikipedia.org/wiki/Keyboard_matrix_circuit

BTW, organ manuals typically have 61 keys (5 octaves plus an extra C).  Modern piano keyboards have 88 keys.