OB-6 Unstable USB MIDI - not only Logic, Windows & Live also

[retango]

Hi,
I see there is another thread reporting midi problems over USB, but the title restricts it only to Logic.. I can confirm I have the exact same problems in Ableton Live (Windows), and even without Live, just monitoring the OB-6 from Windows (Midi-OX). Midi over USB constantly disconnects and reconnects. I can't use the OB6 Plug-SE at all
I'm on OS 1.2, with these settings: Midi Channel All, Param Xmit nr, Param Rcv nr, Midi Clock off, and Clock Port, Midi Sysex and Midi Out USB. I already uninstalled the OB6 midi driver from Windows, and reinstalled it by plugging the cable, but with no changes. If I use Midi ports instead of USB, it works OK.
I also have a P6, and Midi over USB works flawlessly, never had a single problem. Strange since I thought both keyboards would share the same USB Midi hardware..
I will use the Midi connectors in the meantime, but please DSI fix this since I need those PC Midi ports for my older gear without USB midi!
Thanks a lot!

[arhythmetic]

TL;DR On your DAW, try offsetting MIDI output to OB-6 by a tiny ms amount - 0.25ms did the trick for me.

OB-6 desktop v1.2.0
DAW: Reaper 64bit v5.50reapre2
OS: Win 10 Pro 64bit

I contacted DSI support about this - I had also posted in the other thread (the one mostly referring to Logic)

Andy from DSI very quickly replied with a suggestion to "add a 1ms delay in Reaper on the MIDI output". I tried this with a .25ms delay and have had no problems so far. That is, on Reaper I went to Preferences > Audio > MIDI Devices, and on MIDI outputs I changed OB-6 "Offset output to this device by:" to 0.25ms.

Hope this helps.

[musicmaker]

Don't see why MIDI timing would make a difference. OB6 should be MIDI USB standard compliant and be able to accommodate the speed of USB. Does changing to a USB 1.1 interface make a difference ? Not sure why DSI does not take this more serious and steps in, tries to duplicate this  and analyzes this with some DAWs and pinpoints the exact root cause and fix it if it's in the OB6.  Makes me scared to throw a lot of money at such an expensive machine with a defunct USB interface and the problem with the tuning of the oscillators mentioned elsewhere. using the DIN MIDI is NOT a solution.

[DavidDever]

Don't see why MIDI timing would make a difference. OB6 should be MIDI USB standard compliant and be able to accommodate the speed of USB. Does changing to a USB 1.1 interface make a difference ? Not sure why DSI does not take this more serious and steps in, tries to duplicate this  and analyzes this with some DAWs and pinpoints the exact root cause and fix it if it's in the OB6.  Makes me scared to throw a lot of money at such an expensive machine with a defunct USB interface and the problem with the tuning of the oscillators mentioned elsewhere. using the DIN MIDI is NOT a solution.

Class-compliant MIDI endpoints can be problematic with regard to timing, end of story–and it's very much dependent upon the desktop application as well.

[musicmaker]

The problem reported here was USB disconnects. That is serious and should never happen. It's unacceptable specially for music production especially serious when it relates to NRPN/CC or MIDI notes. Timing problems could happen for example by overflow the receive buffer of devices with a poor USB implementation by sending lots of data without pause, like sysex to slow DIN devices, that is why some applications include delay parameters (like MidiOX) and because you never know the data arrived or not its better to go slow. (Except for the reliable Sysex 2-way protocol from Roland which no one else implements).  I also use Linux (which only accept class compliant interfaces) with lots of MIDI  USB devices and have never seen a USB disconnect. A good device MIDI USB implementation is class compliant and does not need the "hacks"  of custom drivers to hide the problems of the device like receive buffer overflows. if the device is too slow to handle USB2 data (which seems here to be the case) , it could tell the USB host that it is a USB 1.1 full speed device or even a standard USB device which is still much faster than DIN MIDI.

[DavidDever]

The problem reported here was USB disconnects. That is serious and should never happen. It's unacceptable specially for music production especially serious when it relates to NRPN/CC or MIDI notes. Timing problems could happen for example by overflow the receive buffer of devices with a poor USB implementation by sending lots of data without pause, like sysex to slow DIN devices, that is why some applications include delay parameters (like MidiOX) and because you never know the data arrived or not its better to go slow. (Except for the reliable Sysex 2-way protocol from Roland which no one else implements).  I also use Linux (which only accept class compliant interfaces) with lots of MIDI  USB devices and have never seen a USB disconnect. A good device MIDI USB implementation is class compliant and does not need the "hacks"  of custom drivers to hide the problems of the device like receive buffer overflows. if the device is too slow to handle USB2 data (which seems here to be the case) , it could tell the USB host that it is a USB 1.1 full speed device or even a standard USB device which is still much faster than DIN MIDI.

https://social.technet.microsoft.com/Forums/windows/en-US/99b81d49-bb70-4e44-94e6-5a6d640b5964/usb-device-keeps-disconnectingconnecting-after-upgrading-to-windows-81?forum=w8itprohardware

The blame here was (likely improperly) placed on power management at the embedded processor, but there are plenty of issues raised with regard to class-compliant USB drivers on Windows, in general.

Blaming the OP's issues on DSI is a bit misplaced; I've written custom drivers for USB endpoints, and you're largely at the mercy of the host OS. Time-sensitive stuff really needs a custom driver IMHO, especially for audio, and certainly for MIDI above and beyond a few twiddled notes.

I agree with you–Linux is a much nicer place to operate (I use it everyday), but unfortunately the ecosystem leaves much to be desired.

[musicmaker]

Thanks for that link. I did write USB UHCI/EHCI/OHCI code before. What I have learned  is that most USB issues are caused by power management (always disable that doing anything with MIDI or audio), having a problematic  hub in between (use one with MTT instead if you really need a hub  - or better get another USB PCI card), using non-Intel southbridge based USB controllers. (VIA and other unstable stuff and the likes), using a USB3 interface or bad USB cables.  I would suggest retango to check those first.

Not blaming DSI here because the root cause has not been identified. But it would be nice if DSI could confirm this is an issue they have seen/reproduced  and if this has anything to do with the OB6 and affect other software environments as well or this is really something unique with this users software/hardware environment.

P.S:I prefer Linux any day over Windows and would never use Windows anymore if it only would support hardware with low latency (AVB, Dante, Thunderbolt). I am terrified of MIDI issues with (synth) hardware because it would mean all money went in the drain. I never buy anything that is not MIDI or USB audio class compliant or does not have a well documented MIDI implementation. I know, that means missing out on many opportunities for buying nice equipment but it also relieves GAS.