GM5 MIDI thru/split help

[eptheca]

Hi,

I am making my sequencer and some synths mobile, by mounting them in a case. I would like as few cables as possible.

The SEQ V4 can run a GM5 module through an OTG adapter, so with only one USB cable out from the SEQ, I can get 4 MIDI OUT port over USB.

I have 8 synths that I want to connect, but none of them have a MIDI THRU port.

They can all receive MIDI on different channels, so 4 OUT should be plenty. 

Q: Would it be possible to "split" the 4 MIDI OUT ports on the GM5 into 2 each? If yes, how?

Can I just connect 2 cables to the same OUT?

Can I use a Hexinverter as a MIDI THRU?

Or do I have to have a full MIDI THRU circuit with a MIDI IN through an Opto, then the Hexinverter?

All suggestions are welcome, maybe this is not the best way to do it at all.

Cheers, Hal

[slo]

Here's a passive solution to what you want to do, 1 x MIDI in port feeds 2 x MIDI thru ports, no power supply.

Midi thru box

 

[latigid on]

If it's only MIDI outs that you need, you gain 4 just by wiring to the correct pins on the F4 disco board/J11E (+resistor), and adding a pull up to +5V for the MIDI- signal. 

Unless your diagram is incomplete and you're already maxed out on MIDI IO and Quad IIC...

Edited December 6, 2016 by latigid on

[eptheca]

Yes the diagram is incomplete, I have 2 x MIDI IO and 2 x Quad IIC connected to the Core already :)

But I was thinking for the mobile solution, that I could get away with mounting the GM5 in the case with the synths, and then I only have to connect the SEQ and the synths with one USB cable.

Do you know how long the cable can be between the Core and the MIDI IO and Quad IIC modules, or if there is a "Line Driver" solution for long cables for this purpose? 

[latigid on]

Technically the MIDI outs are running on UARTs, so it should work within standard MIDI spec. The I2C module should be fairly close to the Core. 

Maybe the problem can be simplified on the other end: one or two DIN MIDI with a thru box on the other end. This depends on how much data you're sending and how important clocks latency etc. are. 

MIDI thrus seem to be done with some sort of CMOS buffer (inverter/NAND etc.). Or you could consider the KissBox and send over ETH.

Edited December 6, 2016 by latigid on

[eptheca]

> Maybe the problem can be simplified on the other end: one or two DIN MIDI with a thru box on the other end.

That's a possibility. I could test that and see what kind of latency etc I get.

Cheers, Hal

[ilmenator]

Latency introduced by the simple inverter schematic as mentioned by latigid on will be nothing to speak of and even difficult to measure. It is basically the gate reaction time times two, which is something you can look up in the data sheets.

[yogi]

In reference to the GM5, you should be fine with: TTL Midi out from GM5 pin into 'HC00 gate(as inverter), it's output to two (or more) other gates (also 'inverters') for two Midi Out DINs. Doesn't have to be 74HC00, just need two 'inverters' per midi out, to buffer. LTC fashion, TTL out ->Buffer (two inverters) ->Midi Out DIN circuit

A "Thru" is connected to the Midi In so it's a split of the TTL level out of the opto. But you want to split a Midi TTL Out so just two-way buffers and DIN circuits.

Yogi

[ilmenator]

Here is a link to something usable from M. Bareille. I think I built this many years ago, so it's verified. Just skip anything left of the first inverter (the one connected to pin 6 of the optocoupler). You'll need one hex-inverter to duplicate two MIDI ports.

Edited December 6, 2016 by ilmenator
fixed a typo

[eptheca]

Thanks guys, I have a try with a hex-inverter after the out pin on the GM5 chip, and then a MIDI Out DIN circuit with 220 ohms resistors.

Hal

[eptheca]

That worked great, thanks :)

Used a  MC74HC04A Hex-inverter I had.

I guess I can use the 3 remaining inverters on this chip for the second port and one more chip for tree and four.

What I'm not sure of is the difference in latency in this, compared to the direct MIDI OUT from the UARTs, but I'm pretty sure I can't hear it.

[ilmenator]

It's definitely more a question of how many MIDI events you push through the connection than of how many inverters / gates you have in there. So, sending out two notes on different MIDI channels but on the same port at the same time will result in more sloppy timing than sending out the same notes on two different ports. However, whether you send these messages through another two gates or not does not matter, even if you had one port with the throughbox and one without (and - by this arrangement - could make a direct comparison).

[yogi]

I believe the only impact of the split is adding two propagation delays on each of the midi Outs. For 'HC logic it's about 10 nS for each gate but I would have to check the datasheet to be exact.

Of course as Ilmenator noted, you will see more timing impact with the traffic/bandwidth on a interface. But I'm not sure you would be able to send two notes on different channels at the same time from a single source. I mean, it could happen if you were merging two midi devices onto one interface, but with just a Seq sending to an interface, UART or USB, it should always be one package at a time regardless of channel. OTOH the GM5 may have more trouble with processing two, three... USB streams in parallel, leading to one UART out of sync with another UART. I've never seen this but I've yet to load a GM5 to this extreme :)

Yogi

[ilmenator]

My point was that propagation delay at gate level is really nothing to worry about. At 5V logic level, propagation delay for an NXP 74HC04 is 7ns per gate at 25°C, i.e. 14ns in total. A 10ms delay between key stroke and sound onset for sounds with short attack times is not noticeable for around 50% of the population (="acceptable"). It is a little less for trained musicians, but we are still talking a factor >500. eptheca, your safety margin is very comfortable here...

[eptheca]

Thanks for the info.

I am also considering another option.

Is it possible to use a "multi midi cable"?

Does MIDI cables need individual shielding, or could I use a D-Sub connector cable like a VGA cable with a DE-15 connector from the SEQ to the case with the 8 synths.

So the 8 M- pins from the 2x MIDI IO modules and 1x Quad IIC module get an individual wire, the Ground goes to the shield of the VGA cable and the 5V gets a wire.

On the other side of the cable, in the synth-case I could connect a MIDI wire and a 220 ohm resistor from the 5V to the MIDI IN on each synth.

Would this work? Or will it be "cross talk" between the "MIDI wires" inside the cable.?

 

[ilmenator]

That would be roughly the same principle as on my MIDIbox Matrix with the BOBs, see here. There is also a schematic in the link you could use. A simple 9-pin serial cable would do the trick, no need for a stiff VGA one.

Edit: no risk of cross talk at all.

Edited December 8, 2016 by ilmenator

[eptheca]

Hey, that's a cool project.

So a 9-pin shielded serial cable, and "BOB's your uncle" ;)

Do you have any suggestions where to get one?

I was looking for one earlier today, but none of them where 1:1 in pins and had wires on all the pins, that's why I thought of the VGA, I have tons of them.

But your right, they are stiff and thick

Hal 

[ilmenator]

I got mine from Reichelt.de where they are very cheap - 3m for 1,95€. If you go for 9-pin extension cables (one male and one female connector) you can be pretty sure that they are wired 1:1, which is not necessarily the case for male-male or female-female cables. PM me if you want to have a BOB PCB, I think I should have a few more of these somewhere.

You could also directly attach the BOB to the STM32F4 core board with the STM32F4-to-BOB adapter board, if that is what you wanted to do - you would then replace the MIDI-I/O module with this combination. But, to be honest, I don't quite remember if I had these boards ever made or if I just did the design for it... I could do some searching over the weekend if you were really interested.

[eptheca]

Thanks for the tip, and the offer.

I think I will still need my Quad IIc for the total of 8 direct MIDI OUT? Or can I get that with the STM32F4-to-BOB adapter board?

I don't have much space in this case I am making, so I am trying to do it with as little modules/components as possible (see image)

It just felt unnecessary to plug in and out 8 MIDI cables each time. A short serial cable will work great I think, also for when I detach the lid from the SEQ part and have them next to each other.    

[ilmenator]

19 hours ago, eptheca said:

I think I will still need my Quad IIc for the total of 8 direct MIDI OUT? Or can I get that with the STM32F4-to-BOB adapter board?

That's true, with the BOB you get 4 in ports and 4 out ports - that is by design, as J11E on the core board provides exactly these signals. If you need 8 out ports you will still have to use the Quad IIC on top.

That is a very nice setup / suitcase design you have there! Just having to connect two single 9pin serial cables (or one 25pin dsub/printer port style cable) is definitely going to be faster to connect :-).  On the other hand, you could probably also just use a 26pin box connector with a flat cable connecting the two halves. That way, you could leave the cable plugged even when you close the lid, and only disconnect it when you want to detach the lid (or leave the cable long enough underneath the front panel).

[eptheca]

Thanks :)

I am hoping to use my music equipment more with this form factor. It's a pelican- style case that I have mounted rails in. 19" rack rails for the SEQ and Eurorack rails for the synths. They are all MIDI synths, no modular analog synths (yet anyway). 3+3+1U, 93HP 

Good ideas about the flat cable :) The less I have to do to set it up, the better.

Is the Ground from MIDI OUT to the shield of the MIDI cable important?

I have to figure out about the powering of the synths, and the audio out. Trying to figure out a power supply with 5V, 9V, 12V and 15V.  I guess I have to accept some cables.