If it's possible to transmit midi over spdif for jitter free transmission, like expert sleepers es-4 demonstration here: Jitter-free MIDI from the ES-4 - YouTube


is it possible to somehow transmit midi over spdif without all the es-4?

hm, I proposed that already 10 years ago since MIDI quickly runs into problems when using more than some channels w/o SYSEX.

I did some projects of electronical development (my primary business) which came up with some similarity to MIDI controlling, in terms breath controlling: the manual movement of a hand had to be recognized, encoded and trasnsmitted making it necessary to sample around 1000 pulses / second to model a movement correctly in 3 axes. With that precision it was even possible to sample a first class pianist's finger movement correctly in order to represent it on a synthesizer, which cannot be done by MIDI because of the low speed.

Well finally I choosed COAX cable with common S/PDIF transceivers to solve this problem, because they are there and available and cheap, because widely used in audio. The customer was fine with it, since it was so cheap and parts easily to buy :-) With a common audio channel over S/PDIF one has 2x24x48kHz data bandwidth and thus roughly >100 times more speed to transport the data.

Later I ported this solution to my DIY audio workstation were I am able to do and program, what I want (FPGA-based, interlinks with LVDS at 384kHz) And this is exactly the problem: It is a single solution. The music gear has to be wdely ready for this and still now, they are going with the old fashioned MIDI interface at mouse speed. I talked to ALESIS and also announced my solution to TASCAM, but there was no interest.
Major Companies will have to start to do so, offering the interface.

AFAIR some years later, Yamaha MLAN an ideas for ethernet based links, came out, which - to me - is exactly the wrong way. Ethernet-Transceicers are much more expensive and not that easy to handle, and have to be >100MHz bidir to overcome a ADAT based 8x48k lane and then it is still jittering (and has more initial latency).

Well let's wait and see, what happens the next 10 years.

If some companies started with it, they could do it very quickly, since most of the gear nowadays have S/PDIF input's physically present. It is only e question of DSP speed to use the existing midi stack.

One word: Jitter will of course also then not be removed, but downsized by ratio 100.

Wow. I almost missed this post! This is a really forward looking perspective on the possibilities of MIDI. Have you tried taking these ideas directly to the MMA (MIDI Manufacturers Associaton)? They're currently in the working on the MIDI 2.0 spec. Your ideas could be both welcome and timely.

Maybe the tack should be starting with getting one person of influence interested. There's too many lacklustre and dispassionate dullards at the front desks of most of these big companies... You need to talk to someone who "gets it".

Midi 2.0 is that "midi hd" development than been abandoned?

It doesn't look like it's abandoned: according to this page at the MMA's site, as of a few weeks ago MMA working group on MIDI HD is still at it and demoing the technology hoping to get manufacturers to jump in. But it does look like they're using Ethernet as the transport.

That's the point. Using Ethernet has two issues:

1) It comes with additional transport dynamics causing jitter, so time stamp orientated processing is the only chance. Realtime s impossible, or let's say was no improvement.

This time based processing of course is possible / suitable and most probably the best for nowadays midi processing via computer, with external MIDI-processors on adjecent PCs with e.g. a soft samper. On the other hand: Direct processing from the virtual DAW like Cubase into a softsampler / VST shuns that problem anyway.

2) There is no gear for it. Every device has to be equipped with ethernet capable chips which extends cost and demands, because processing larger Ethernet Streams requires CPU power especially when done in little tiny MCU's used in embedded devices.

High Speed Duplex Data Streams known from GigE like used for Video (GigE-Vision) can only be done with FPGAs. I am ready for this since I do that regularly for my primary engineering business and my workstation is ready too.

But again I would like to focus on an easy entry step of this technologie:

Old gear, or better let's say, "exsiting , current" devices, DO have the S/PDIF interface already and it was a small step to implement this rapidly. You just have one input less.

With the widely spread ADAT-Interfaces, It was even easier: 7 channels left for audio.

> Have you tried taking these ideas directly to the MMA ?
Yes, I applied with a prepared paper - I seemed to be received but not accepted as a proposal for a convention / congress or such. I also checked AES for this and had been aswered that this idea was already known.

Well, I assume, many persons had come to this idea already since it appears logical to me to enhance the MIDI interface in speed once having understood, how slow it is when transporting some more channels and SYSEX over it.

With MIDI over S/PDIF I interlink several FPGA boards locally with only 0,1ms latency = maximum jitter for 1023 MIDI-channels and I use a 5 Byte MIDI and 1023 values (10 Bit) instead of only 127 for channels, parameters and notes. (127 is much too coarse to my mind).

Technically this is a piece of cake, since even $5-FPGA-devices offer LVDS and what you only need is a transformer or a balun OP. See the image: "MIDI Routing" here: http://www.96khz.org/cyclone3platform.html

For Ethernet, you need a chip. You could integrate the MAC features of course into an FPGA and use only the PHY, but for microcontroller based operations (like in the most music devices) you need a MAc too. Two more chips instead of two more pins at a DSP.

If you do not have any ethernet-compatible devices, must use an adapter with a MIDI-Lan receiver transforming back to MIDI, and then you have it again, the latency of the transmitted bytes over the slow serial connection, even if it is only 5 inches :-)

How come multi channel audio, (much more larger data packets than midi) gets transmitted over usb without any problems, and midi with this tiny data gets into so much trouble???

USB is of course another idea, but 10 years ago there was no (quick USB) so I had expected though to have changed to another standard already since MIDI is slow.

Well, looking at the current situation, USB 2.0 is established and easy to use, since you can drive up to 10m distance without repeater using good cables and at least 3-5m with standard cables. USB 2.0 offers 480 Mbit of bandwidth, de facto 60kB/s which is appropriate for 20000 3Byte Midi messages. Assuming real time processing for live recordings while playing 6 finger chords + Controllers will cause maybe 10 MIDI events and a transmission delay of around 0,5ms for the last event in comparison to the first, keeping in mind, that all of them should be at the SAME time. So this is acceptable.

But:

a) The problem of low modulation options persists: If one likes to transmit detailed information from breath controllers with more channels, where velocity and volume are represented with more midi events, the delay quickly increases up to 2-3 ms which is definetely not "the same" point of time" anymore. It will cause unintend and unpredictable jitter.

b) not all the devices do have / will have USB, since USB communication causes issues when performed with PCs, additional offset delay occurs anyway.

c) Assuming, one uses USB, I'd suddenly like to ask the question, "why Ethernet then" ? Where is the progression? GigE has the double bandwidth only when using the around 850 MBit of Data, what GigaBit effectively has, disregarding packeting small packeges having much realtive overhead. Also the PCs usually run into problems with more than 800Mbit. If one takes PC usage into account, GigE is no advantage. One has to spent an additional NIC containing PHY+MAC chips, since most PCs do not have 2 NICs and running GigE Music over switches will be very funny. USB ports are available in masses. Also when disregarding PC usage, the implementation of USB in new msucial devices is much easier since the chips are much cheaper and easier to handle. It is basically a RAM-access protocol (write data the pipelined way, and that's it). Ethernet Communications is much more complicated and furthermore, there is currently one one MAC chip out in the world which offers RAM-like communications, (ASIX 88xxx) and this chip is slow (only 400 MBit in bw maximum) and cannot do dual way communications to the PHY. The benefits of Ethernet (TCP/IP ready with data security cannot be used with real time MIDI).

Conclusion: Ethernet does not increase the options or then bandwidth significantly in comparison to USB. If people need more distance in between their devices, they can use a USB repeater. If they need to move data to other locations, USB2ETHERNET-converters are present and cheap for the low number of people who really need that.

In the future, USB 3.0 will overcome the GigE, since 5GBit will be possible over short distances.

Ethernet for MIDI is a dead horse!

just as an update: I am creating a version of my synth also for B2C use and a midi mix matrix is built in. I intend to make kits available for DIY to built a synth abd so midi transmission the above way is possible with that.

Speaking of sending midi over s/pdif coax, and this may be getting off-track from original topic. But I've done this through Cubase's own proprietary VST System Link for a while now, but between computers, not directly among midi device/sound module etc. System Link can send midi, clock, synchronization, transport and digital stereo audio, all down the same s/pdif coax without noticing issues.

This is old technology by now, I would have thought I would have seen this being done in many other DAW's other than Cubase, or in 3rd party software apps giving System Link-like capability (and not just midi capability), but I have not.

All my audio interfaces have coaxial s/pdif I/O's like many interfaces commonly have, coax cables are fairly cheap. I bought a used, inexpensive and widely available matrix switcher (a Shinybow SB-5588), which was actually primarily designed for video purpose situations, but it has analog & digital s/pdif I/O. I only use the s/pdif digital audio I/O's. The matrix switcher allows me to properly, and evenly distribute all of the above mentioned VST System Link data from any one computer, to all other computers simultaneously...including midi data.

Not to forget, this is / was an easy method to integrate HW synths into DAWs in real time easily without the asynchronous issues caused by USB.