USB MIDI slower than plain MIDI on Fantom X

[Arjan]

Hi,

Has anybody else noticed that the USB MIDI interface on the X is slower than the regular MIDI interface? I discovered this while testing my Fantom X Live Commander software with the USB MIDI interface. Performance was very much inferior compared to connecting the Fantom to my PC using an Edirol UM2-EX USB MIDI interface.

Can anyone confirm my findings? Of course you don't notice this if you play just a few notes on one MIDI channel but if you start layering a few channels it becomes noticable very quickly.

I recently got the Fantom X Service from a nice fellow clan member so I looked up the block diagram to see if there could be an explanation for this and it seems there is.

According to the block diagram on page 43, the physical MIDI ports are hardwired directly to the main CPU while the USB interface connects to an internal bus which also connects the LCD and the WX chip which I guess is the actual sound engine. I suppose that the LCD especially may be taking up so much bandwidth that it may interfere with the real-time MIDI stream, although I'm not sure how much traffic is going on if the LCD content is not changing much.

Any other explanations?

[Arjan]

Bump. Anyone?

[karlosserio]

Maybe because the interface is USB 1 and not 2?

[Arjan]

Thanks for your idea but I doubt that raw transmission speed is the reason. MIDI is only 31.25kbaud (more or less similar to 31.25kps minus start/stop bits) whereas even the slowest USB 1 speed is 1.5Mbps or 1500kbps and normal 1.1 'Full speed' is 12Mbps. USB 2.0 is 480Mbps. I'm not sure what standard the X uses, I thought it was USB 2.0 but I could be wrong.

You can't really compare these numbers of course since MIDI is a point-to-point uni-directional interface (only one sender and one received on the same wire) whereas USB is a bus-based protocol with multiple senders and receivers (although there's only one initiator).

[exileinoblivion]

Have you gotten any further with this? Id be really interested in hearing about your findings

[compusic]

I have heard of quite a lot complains about the timing of USB-MIDI.

And I got the article below casually.

http://www.digitalprosound.com/2001/01_jan/features/mid...

Interesting, though it's a little old. Do not know the situation now, and can't find relevant info on the net.

[Arjan]

Have you gotten any further with this? Id be really interested in hearing about your findings

Not really and in fact there's not much I can do exept ask other if they can confirm my observation.

I did try another MIDI interface, a MOTU Express 128. It did perform much better except that occassionally it would suffer very serious delays and within minutes it would also get stuck notes (meaning that a note off message had not made it to the synth).

You might think that this could be a DPC latency issue (some other driver on my system taking to much processing time during interrupts) but my Edirol UM-2EX performs perfectly on the same system with no noticable delay whatsoever, no stuck notes and overall flawless performance.

I can only conclude that neither the Fantom driver nor the MOTU driver have been implemented as efficiently robustly as the Edirol. I guess the Fantom doesn't have kernel streaming drivers or something (which if I'm correct is what Rolands FPT 'technology' is mostly about).

And the MOTU? I have no idea and some people claim that it performs flawlessly on their systems but when I installed it I noticed that it installs no less than three drivers just to access on MIDI interface, including a virtual USB port driver. Sounds like inefficient legacy compatibilty layers to me which may work fine on a powerhouse machine but on my lowly Netbook with Atom processor it may take too much CPU time. Or Roland/Edirol really did something very different and clever in their driver design. Too bad they only did that for the Edirol interfaces and not for the Fantom itself.

[Arjan]

I have heard of quite a lot complains about the timing of USB-MIDI.
And I got the article below casually.

Note that it talks about the 'upcoming USB 2.0' standard', ie this is an old article.

It's still somewhat true that USB is not perfect for real-time performance but it's really not that bad, especially for MIDI. I believe modern USB MIDI interfaces have latencies MIDI interface is located: in a separate box or inside the Fantom but essentially it's the same thing, except for the communication path inside the Fantom where the MIDI data must share a general purpose bus which is also used for the display among other things. They seem to use completely different drivers though and indeed Roland makes no claims regarding FTP being used in the Fantom MIDI-USB implementation.

One more thing, MOTU's MTS (Midi Time Stamping) technology is only useful if you are playing back pre-recorded MIDI data from a sequencer. It cannot do anything to accelerate and stabilize timing of anything you actually play live, with your hands, in real-time. MTS simply sends data to the USB MIDI interface ahead of time along with a timestamp that tells the interface at what point in time it should actually transmit the MIDI message to the actual MIDI output port.