Gear used:

3 MIDI sources: 486 DX2/66 PC, vintage Macintosh through a midi interface box, regular synth midi controller

3 MIDI destinations: Roland MT-33, Roland Sound Canvas 88ST, Kawai LG-Mega

All of these are known to work, I have used them extensively under every combination.

Goal: to make a functional 3 in, 3 out switch that selects one in and one out at any given time.

Gotcha: I can't use the regular IN->THRU method of chaining the destination devices together because only my Roland MT-32 has that port and none of my other devices have it.

I made a perfboard that connects pins 4 and 5 of the midi sockets together. I use two double pole, quadruple throw (one throw is redundant under this solution) switches. As for the central GND pin, I simply soldered a path and some wires to connect all 6 devices together, which is where the potential problem might arise from.

i have tested every single one of the 9 possibilities with most of my devices and the wiring is sound, so the perf board is properly verified to work one path at a time.

Now, onto the problem. If I leave all the devices connected, and just the one source powered and the one destination powered, the signal doesn't go through. The scenario I'm most interested in is to have my Mac SE/30 running Cubase and sending midi signals through its modem serial port, through a midi interface box, from which comes an OUT midi cable, into my 3x3 switch, followed up with the cable that goes to my Roland SC88ST.
If the cable that goes from my interface to my switch is disconnected, I see my midi interface light up with activity consistent with the music. If I reconnect that cable to re-establish the midi path, the activity LED dims a little and stays static (no signal gets through).
If I then disconnect every unseeded devices from my 3x3 switch but only keep the 2 strictly necessary ones, it works again.

Possible explanations: 1) the various GNDs interfere together and "drown" the signal strength somehow 2) the aging power supply of my Mac SE/30 is marginally not powerful enough to push through the totally wired up switchbox 3) these 6 gnds shouldnt be wired together?

I have a scope to diagnose the difference between different connection scenarios and could post pics in followups but I would greatly appreciate some guidance to zero in on the wrong assumptions I might have about this project.

 

There may be outputs that do not default to high impedance in one or more of the modules. That would result in shunting the signal to ???, which would impair the function. So the safest choice is a multi-pole three position switch, switching even the ground connections.

 

I am confused....

I always thought that MIDI data used pins 4 and 5 only (of 5-pin DIN plug/skt). Ground is only for cable shield.

I would switch pins 4 and 5. That is the sensible way to do this. (better still switch 2, 4 and 5 with 3 pole - 3 way switches)
Maybe you might get away with connecting all pin 4's together and switching pin 5. I could not live with such a terrible bodge, but maybe you can.

I would leave ground pins unconnected.

That is my understanding, but I could be wrong. Not tried it that way for myself.

Pin 4 is not Vcc, it is data+ (which is usually connected to Vcc, via resistor)
Pin 5 is data-

Pin 2 is ground, but should be connected at transmitter end to cable shield, if required.

 

What type of slide switch? do you have a mfgr and part number?
Pin 2 should not be connected and left unconnected.
P5 of "midi out" is the midi logic signal, and connects to the "midi in" internal opto isolator led cathode.
P4 of "midi out" is connected to 5v (or 3.3v depending on the midi systems used) via a resistor and supplies voltage to the "midi in" internal opto isolator LED anode also via a resistor. Since your "box" is not designed to midi specifications, VCC is probably being overloaded and unable to drive the opto's.

 

I always thought that MIDI data used pins 4 and 5 only (of 5-pin DIN plug/skt). Ground is only for cable shield.

[...]

I would leave ground pins unconnected.

Yes, you're right. I misread the MIDI schematics on that day I soldered that switchbox.

What type of slide switch? do you have a mfgr and part number?
Pin 2 should not be connected and left unconnected.
P5 of "midi out" is the midi logic signal, and connects to the "midi in" internal opto isolator led cathode.
P4 of "midi out" is connected to 5v (or 3.3v depending on the midi systems used) via a resistor and supplies voltage to the "midi in" internal opto isolator LED anode also via a resistor. Since your "box" is not designed to midi specifications, VCC is probably being overloaded and unable to drive the opto's.

The switch: G-1138S-1110
SWITCH SLIDE DP4T 11A 125V

https://www.digikey.ca/en/products/...LrPLV_OT3IWn1MARh1vHbLxr7RdQ8NSRoCOQAQAvD_BwE

thanks for the sanity check. You're right about the pins as well. It's true that MIDI Ins have optos, I have done another perfboard for a mt-32pi project where I used a H11 and the schematics above. In this case, however, I expect every destination used here, commercial MIDI sound module boxes, to be well behaved internally, so no need for something above wires and a switch.


Thanks, it all works after desoldering my GND bridges!

 

Glad you succeeded

Pleased that you are switching both data lines - that should work well.

 

Here's a video I did of the whole thing: