Just discovered that nasty MIDI bug. Ouch! :-(

General questions about the fusion belong here.

Moderators: parametric, Jesse, jeebustrain

Just discovered that nasty MIDI bug. Ouch! :-(

Postby 2ManyKeys » Tue May 22, 2018 6:04 pm

For basically the first time since I got it I was controlling another module with the Fusion last night. I was playing along when all of a sudden my Blofeld seemed to go berserk! I thought the little Waldorf was toast and then realized it was actually strange delayed MIDI coming out of the Fusion! At first I wondered if my Fusion was damaged during my RAM install but realized that the MIDI is all handled by the Coldfire CPU which was obviously working fine controlling everything else. My mod wheel usage seemed to trigger the problem so I wondered if it was getting dirty and even put it through another calibration and test without issue.

Needless to say I got on the forum and started searching and of course came upon this thread.
http://promusicproducts.com/forum/phpBB ... 3&start=60

I have to admit this is rather disappointing. Does tweaking the clock settings get the Fusion to the point where it can control external gear and remain stable? More importantly, can the Fusion be controlled externally by a sequencer without going nuts?

You guys are years ahead of me in this area. How have you been able to hook the Fusion into your setup successfully?

Thanks,

Scott
2ManyKeys
Advanced
Advanced
 
Posts: 181
Joined: Sun Feb 11, 2018 10:55 pm

Re: Just discovered that nasty MIDI bug. Ouch! :-(

Postby 2ManyKeys » Tue May 22, 2018 8:48 pm

Okay so I tested it with a few imported MIDI files. At first the Fusion was doing it's crazy delayed MIDI output thing after a few plays of the Start Trek Voyager theme. Oddly, it sounded like most of the tracks were still playing correctly though. Maybe there are just certain tracks or channels that go wacky?

Then I turned off all MIDI clock and sequence control options. That seemed to take care of that. Of course controlling other gear without MIDI clock is a problem.

Chris sent me this crazy piece by Bartok. 29 tracks and virtually continuous notes. Pegged the MIDI activity indicator on my JV-1010 but the Fusion remained stable through a number of plays through of that very long, 630 bar piece! The Fusion even handled the very frequent tempo changes in the piece without issue.

Now I need to test if the Fusion can handle being controlled by a sequencer. I really hope it can handle notes and incoming clock without freaking out.
2ManyKeys
Advanced
Advanced
 
Posts: 181
Joined: Sun Feb 11, 2018 10:55 pm

Re: Just discovered that nasty MIDI bug. Ouch! :-(

Postby EXer » Wed May 23, 2018 4:59 am

2ManyKeys wrote:I was playing along when all of a sudden my Blofeld seemed to go berserk! I thought the little Waldorf was toast and then realized it was actually strange delayed MIDI coming out of the Fusion!

Are you sure it was the fault of your Fusion?

Was the "> Global MIDI > Clock" parameter set to "Internal" or to "Auto" on your Blofeld?

If set to "Internal" the Blofeld uses its internal MIDI clock, which can create conflicts when it receives MIDI clock information from an external device, hence the going berserk.

If you want to use the Fusion as a master keyboard for the Blofeld you should keep the transmission of the MIDI clock active on the Fusion and set that parameter to "Auto" on the Blofeld so that it deactivates its internal clock when it receives a clock signal on its MIDI port.

► More generally, the MIDI clock should be set to "transmit on" on the master controller and it should be set to "external" on the controlled devices (except in specific cases).
Last edited by EXer on Wed May 23, 2018 5:38 am, edited 2 times in total.
8HD + 2 x Sector101 EXM-E3 128MB DRAM modules
User avatar
EXer
General
General
 
Posts: 85
Joined: Sat May 02, 2015 11:42 am
Location: France

Re: Just discovered that nasty MIDI bug. Ouch! :-(

Postby 2ManyKeys » Wed May 23, 2018 5:13 am

Definitely the Fusion. Did the same thing with the JV-1010. The problem seems to arise when MIDI data of a certain density is being pushed out of the Fusion along with MIDI clock. I was able to create the problem fairly easily while using the mod wheel while the Fusion was sending but not receiving clock. Controllers like the mod wheel create very dense streams of CC messages. I think something in the internal logic of the Fusion breaks when it can't find a time slot to insert a clock message. At that point it starts delaying normal MIDI messages in some queue in order to insert the clock message on time. That is probably only supposed to be mS of delay but has a serious bug.

In the design of MIDI devices, receiving a MIDI stream is relatively easy. The stream is a realtime sequence of events and the MIDI device simply has to react to an already properly sequenced stream. Generating a stream is a harder problem. MIDI, over the classic DIN connectors, cranks along at the laughably slow rate of 31250 baud. That means that quite frequently the exact timing of MIDI events cannot be maintained. Even big chords will create a situation where notes are being held in a queue and inserted with slight delays because the time slot for the current note event has already been filled. As the information density increases, the length of the queue gets longer and so do the delays. Data also has to be prioritized by type. A mod wheel can generate very dense MIDI event streams and if not handled properly in the queue mechanism notes would be audibly delayed by less important CC events. Instead the device should put the note events in higher priority and make sure their delivery is as accurate as possible, then inserting CC events where there are open slots. This gets all that much more difficult when adding clock messages. Now you have clock messages which should be highest priority and are fairly dense compared to notes. Add some mod wheel activity and things can get difficult quickly.

Turning off send clock seems to fix the problem in the Fusion.

I'll put money on it that the Fusion keeps increasing it's queue length until it hits some error condition that isn't handled and then the long TX queue/delay remains in force until the Fusion is power cycled.

This delay may not be the same for each MIDI channel. It may only be on the "global" channel. I'm not sure about that but when I introduced the problem while complex multi-track songs from the sequencer, it sounded like many of the notes were still properly timed and some others weren't. Maybe notes on the same channel as the clock events?
2ManyKeys
Advanced
Advanced
 
Posts: 181
Joined: Sun Feb 11, 2018 10:55 pm

Re: Just discovered that nasty MIDI bug. Ouch! :-(

Postby 2ManyKeys » Wed May 23, 2018 9:26 am

I wonder if this problem would go away as long as note and CC events were only transmitted on a channel other than the "global" channel where the clock events will be placed. This might be a workaround at least for Song mode.

Hmm... have to test this.
2ManyKeys
Advanced
Advanced
 
Posts: 181
Joined: Sun Feb 11, 2018 10:55 pm

Re: Just discovered that nasty MIDI bug. Ouch! :-(

Postby EXer » Wed May 23, 2018 10:29 am

I don't understand what's going on with your Fusion. I've never experienced any problem using my 8HD as a master controller with my EX5R or my V-Synth XT, and I have just tested it with my Blofeld.

I just can't make my Blofeld go berserk, even with "Transmit MIDI clock" checked on my Fusion, with "Clock" set to "Internal" on my Blofeld (which should conflict with the clock received from the Fusion through the MIDI port) and with heavy use of the MOD wheel on an "Arp" category patch on the Blofeld.
8HD + 2 x Sector101 EXM-E3 128MB DRAM modules
User avatar
EXer
General
General
 
Posts: 85
Joined: Sat May 02, 2015 11:42 am
Location: France

Re: Just discovered that nasty MIDI bug. Ouch! :-(

Postby 2ManyKeys » Wed May 23, 2018 12:12 pm

That's interesting as our setups are virtually identical with the exception of year. If you go into diagnostics mode and check the board hardware revisions what do you get? Mine are revision B.

According to the thread above and others, I'm far from the only one who has seen this same problem. Chris? Jesse? How about you guys?
2ManyKeys
Advanced
Advanced
 
Posts: 181
Joined: Sun Feb 11, 2018 10:55 pm

Re: Just discovered that nasty MIDI bug. Ouch! :-(

Postby EXer » Wed May 23, 2018 12:40 pm

2ManyKeys wrote:If you go into diagnostics mode and check the board hardware revisions what do you get?

{B}

___

Blofeld OS 1.23 (early versions are known to have has MIDI timing issues).

AFAIK latest OS currently available from Waldorf is 1.22 (1.23 introduced optimization of sample memory managing with Spectre, but Waldorf pulled it because it was not fool-proof).
8HD + 2 x Sector101 EXM-E3 128MB DRAM modules
User avatar
EXer
General
General
 
Posts: 85
Joined: Sat May 02, 2015 11:42 am
Location: France

Re: Just discovered that nasty MIDI bug. Ouch! :-(

Postby 2ManyKeys » Wed May 23, 2018 2:00 pm

Like I said I'm not worried about the Blofeld. I produced the same issue on the Roland JV-1010 just as easily.

So we have the same hardware.
2ManyKeys
Advanced
Advanced
 
Posts: 181
Joined: Sun Feb 11, 2018 10:55 pm

Re: Just discovered that nasty MIDI bug. Ouch! :-(

Postby 2ManyKeys » Fri May 25, 2018 7:56 am

It's been a busy week so I have had little time to test this. I tried it again the other night with Clock Tx on and couldn't reproduce it as easily. However I had both the MIDI Sequence control options turned off as well.

I will test this more when I get a chance and see if I can nail it down.
2ManyKeys
Advanced
Advanced
 
Posts: 181
Joined: Sun Feb 11, 2018 10:55 pm

Re: Just discovered that nasty MIDI bug. Ouch! :-(

Postby 2ManyKeys » Sat May 26, 2018 7:02 am

Yeah, it's the MIDI Clock TX Out that causes the problem. I'm going with my theory I detailed above. Relatively sparse MIDI data won't be enough to trigger this queue buffering error. However something like the crazy Bartok MIDI file I imported into the Fusion does it. It's 29 tracks and 630 measures of arpeggios, trills, and runs galore in 2/4 time. With MIDI Clock Tx On Bartok had my Fusion going nuts within a few minutes. It's easy to test. Once the buffering has gone into broken mode you can press stop on the sequence and strange slow extended notes will continue to play and if you have a module connected with a MIDI in activity LED it will continue to light up. Again this is only a problem for external synths. The Fusion's internal voices never go wonky.

Without MIDI Tx On, I played the Bartok for probably an hour without the problem.

I turned the volume down during most of that. It's not quite atonal, but it doesn't follow normal harmonic progressions and at least for me is not a pleasant listening experience. Great test file though.

None of these tests were done with MIDI Clock RX on.

The Bartok file posted by parametric can be found in this thread.

viewtopic.php?f=4&t=6933&p=36695#p36695
Last edited by 2ManyKeys on Mon May 28, 2018 8:58 am, edited 1 time in total.
2ManyKeys
Advanced
Advanced
 
Posts: 181
Joined: Sun Feb 11, 2018 10:55 pm

Re: Just discovered that nasty MIDI bug. Ouch! :-(

Postby parametric » Sat May 26, 2018 10:02 am

2ManyKeys wrote:With MIDI Clock Tx On Bartok had my Fusion going nuts within a few minutes. It's easy to test. Once the buffering has gone into broken mode you can press stop on the sequence and strange slow extended notes will continue to play and if you have a module connected with a MIDI in activity LED it will continue to light up. Again this is only a problem for external synths. The Fusion's internal voices never go wonky.


Just a thought,

What Midi I/F are you using to drive external modules?

If it's one of these, then they are famously useless at handling LARGE amounts of midi data.

SYSEX DUMPS are a task they generally fail at . .

cheap midi interface.JPG
cheap midi interface.JPG (17.64 KiB) Viewed 186 times


ANYTHING slight higher up the tree i.e. with a "Name" will generally be NO PROBLEM . . . .

parametric
Alesis Fusions 8SSD + 6SSD/384Mb/120Gb SSD/Yamaha SY85/8.5Mb/1024kb/Yamaha MOXF6/1024Mb Flash/Akai ADVANCE61/NI Komplete11 Ultimate/Roland MT32/Bachmann double overstrung Baby Grand Piano/Win10 Pro/Ubuntu Mate 15.04/M-Audio MicroTrack

https://chrisnmiller.co.uk/Chris
User avatar
parametric
Extreme
Extreme
 
Posts: 620
Joined: Mon Aug 02, 2010 7:22 am
Location: Watford UK

Re: Just discovered that nasty MIDI bug. Ouch! :-(

Postby 2ManyKeys » Sat May 26, 2018 12:17 pm

Nope, I don't use that stuff. Only good interfaces and routers like MOTU. However, this is directly out of the Fusion so it has nothing to do with the device receiving the MIDI out from the Fusion. It is still not consistently easy to replicate, but it never happens with MIDI clock Tx off. Once the Fusion gets flakey, a power cycle is the only fix. Try that Bartok file yourself. Set it to loop the whole song with Clock Tx on. I'll bet that if left on for a while it will finally go flakey on the MIDI out.
2ManyKeys
Advanced
Advanced
 
Posts: 181
Joined: Sun Feb 11, 2018 10:55 pm

Re: Just discovered that nasty MIDI bug. Ouch! :-(

Postby 2ManyKeys » Sun May 27, 2018 1:38 pm

I spent some more time testing this and in my tests, the MIDI Clock TX is always the problem. The problem does not manifest itself in a simple repeatable manner which is why Alesis didn't catch it in their QA tests. It does seem to be related to the density of the MIDI stream combined with the Fusion trying to insert MIDI clock events. I have never seen it happen with Clock TX off. I switched off the midi router output to the Fusion and then even disconnected the MIDI IN cable from the Fusion.

The Bartok file noted above is very very dense - and imo not very enjoyable. I don't have too many extra banks so the 29 track Bartok will just pick voices based on the organization of my banks. Anyhow, that Bartok really hammers the Fusion. Both voice engines are hitting between 80 and 90 percent load at times!

If left running with Bartok set to loop and MIDI Clock TX on, eventually the long delayed notes start and the Fusion requires a power cycle.

I imagine for many normal uses with naturally more sparse midi data you can get away with sequencing to external modules with clock tx on without trouble. But it's good to be aware of the problem. I have never seen it with Clock RX only so controlling the Fusion from external equipment shouldn't be a problem in any case.
2ManyKeys
Advanced
Advanced
 
Posts: 181
Joined: Sun Feb 11, 2018 10:55 pm

Re: Just discovered that nasty MIDI bug. Ouch! :-(

Postby mstsfreak » Mon May 28, 2018 4:35 am

From the other topic:

I know this topic about the MIDI-clock is very old and i read it again after all those years, with interest, but i think there was a misunderstanding about this function in the Fusion. Workstations with a sequencer have an option where you can choose if the workstation is the master MIDI-clock or if it syncs to an outside MIDI-clock.
In the Fusion settings, you can choose both options, it syncs to an outside MIDI-clock, but it also sends MIDI-clock (i.e. it is master and slave at the same time). This isn't possible of course, hence the problems. If there had been a 1.25 OS update, this problem would have easily been solved by replacing the two checkboxes by only one, switching from Master to Slave or vice versa.
In OS 1.24 the problem is gone by checking none of the boxes, or only one of the two, i.e. choosing Master or Slave.
Transmit ON, Recieve OFF = MASTER
Transmit OFF, Recieve ON = SLAVE

I rest my case... :shock:
User avatar
mstsfreak
Advanced
Advanced
 
Posts: 401
Joined: Wed Apr 05, 2006 12:14 pm
Location: The Netherlands

Next

Return to General

Who is online

Users browsing this forum: No registered users and 4 guests