fbpx
Skip to main content

MIDI Forum

Notifications
Clear all

Repairing MIDIs

29 Posts
7 Users
0 Reactions
23.4 K Views
Geoff
Posts: 1047
Noble Member
 

Angus,

Did I say it sounded 'right'? That was stupid of me. I very much doubt even my best version sounded 'right' (I assume - in comparison to the original game).

You also say that none of the versions You've been able to play sound 'right' either. But you don't say what was wrong with them, maybe a lot of little things which might be expected that you're NOT using an original SB16 card.

Oh, I can record the Yamaha MU90r version, I'll try and do that.

Geoff

 
Posted : 30/05/2018 8:40 am
Eddie Lotter
Posts: 295
Reputable Member
 

Listening to some of those MP3 files, it sounds like a Roland MT-32. That would explain why only MIDI channels 2 through 10 are used in the MIDI files.

Cheers
Eddie

 
Posted : 30/05/2018 7:07 pm
JohnG
Posts: 227
 

A useful comment, Eddie.
I think you've hit the nail on the head.

I've done some further research and it would appear that Wing Commander was written to take advantage of the "older" version of the MT32.
This also explains why the MIDI file sounds odd played on a GM module/VST.
I found a list of the voices the MT32 supported and which MIDI program number they were invoked by.
The original document can be found by following the links on the MT32 page in Wikipedia.
The details can be found at the URL below.

JohnG.

 
Posted : 31/05/2018 1:40 am
Angus
Posts: 8
Active Member
Topic starter
 

Geoff,

Maybe you didn't use the word "right", but from the way you were talking, it sounded like you weren't hearing any of the problems I was. Anyone who heard what I've heard would know something was wrong, even if unfamiliar w/the original score.

I've listed a few ways in which I've experienced or heard of the failure on the various platforms, but the most up-to-date list is, at 3:55 to the end:

  • Cacophony, like multiple renderings of the same piece or left and right out of sync. (Improved w/the removal of the sustains, but not eliminated)
  • Scratchiness, like a fading AM radio station
  • Silence
  • A minute and a half of droning (eliminated by the removal of the sustains).
  • Sekaiju's playback was also unique, in that there was utter silence after 3:55, with the brief exception of these drums, which I'd never heard on other platforms.
 
Posted : 31/05/2018 5:50 am
Geoff
Posts: 1047
Noble Member
 

I've still got my Roland LAPC-I card active on my old system. That is in effect a MT32. I'd not tried playing through that, as I'd interpreted early comments to relate the SB16 card to GM, and the GM playback seemed OK.

I'll get to try the playback of the original file via the Yamaha MU90r, and via the Roland card (as per MT32) and record both, and post the result.

Geoff

 
Posted : 31/05/2018 3:45 pm
Geoff
Posts: 1047
Noble Member
 

Hello again,

I've just played the file both ways, using the LAPC-I card (MT-32 sound set) and to the Yamaha (GM).

Not really THAT much difference, but there ARE certainly differences.

I certainly could NOT say that one is clearly 'right' and one is clearly 'wrong', but then I've not heard the original version. As in, played from the actual game. I don't know that the version in the .MP3 is 'right' - is it?

I'll get the two versions saved as .MP3. Not done yet.

The sound quality (generally) from the Yamaha is clearly better, but then it ought to be, it's later technology. Some of the instruments are clearly the same, i.e. piano, but various other instruments are similar patch ## in both sound sets. Both versions hit a problem about 4:25 - the Roland card just goes to nothing, the Yamaha goes to a drone on Ch 2 & 3 until the end (5:26).

Hmm - for what it's worth, the Yamaha MU90r supports a C/M sound set, which is basically MT32. Never tried this. I suppose it'll be MT32 with the later level of sound quality. It's somewhat 'enhanced' over the normal MT32, as there's the standard MT soundset for Ch 1-9, and various extra sounds for Ch 11 to 16. I suppose there was never much point in my trying this as I have the 'Real Thing' readily to hand!!

Just tried the C/M version. It works. Not really sure how it compares to the other two versions. When I do the MP3s, I'll do all three versions?

Geoff

 
Posted : 02/06/2018 8:41 am
Angus
Posts: 8
Active Member
Topic starter
 

How about this: since there seems to be a lot of ambiguity over what the MIDI is supposed to sounds like and not sound like, maybe I should upload some samples of what I'm hearing, at the part from where it gets bad, and that might make it easier to decide what to share.

  • WC2-48-cacophony.mp3: no one could mistake this as being correct, unless he though it was supposed to be some hokey psychedelia.
  • WC2-48-removed_sustains.mp3: this is what I got after removing the sustains using Sekaiju, according to above advice. It's not nearly as bad, but trust me, it's not correct, like something's trying to overplay it. Also, notice the long silence at the end. Actually, don't trust me...
  • WC2-45.mp3: this is what it should sound like. It's the original that was used to make the medley. It goes on for well over a minute, which would just about fill the remaining time in WC2-48
 
Posted : 03/06/2018 8:01 pm
JohnG
Posts: 227
 

Angus, Geoff,

The only way you're going to get the same output is by using the same device as this was written for, as it was clearly NOT written for General MIDI (GM) playback.
Or if you can find a device/program/VSTi that will emulate an MT32.

It can be adapted by using the list of voices I referred to earlier by doing the following.
First, in Sekaiju go into List View and tick all the tracks to be displayed (16).
Then select that only Program Changes are displayed.
See the attached image.

Now, using the Roland MT-32 list of voices, match the voices using GM sounds.
That is, substitute by changing the Program Number for the correct sounding voice.
So, the first entry is on MIDI channel 2, right at the start (00001:01:000) reads "48-Strings", but in the MT-32 list of voices 48 is "Square Wave" not Strings.
So, find the GM Program Number (or whatever sound device your using) and substitute maybe GM 81-Square Lead.
The 2nd entry reads "116-Taiko", but the MT-32 voice for Program Change 116 is "Elec Perc 1".
You need to figure out what GM sound to substitute for that and enter that as the new Program Change number.
And so on.
Finding a suitable substitute may be difficult in some cases, e.g.what is Bellsinger? I haven't a clue.

If you can substitute all the MT-32 Program Change voices for GM ones, you may get a fair approximation of the original.

The file continues playing for some time after all Note On messages have occured, in measure 112, because there are two further Tempo Change messages, one in measure 131 and another in measure 144.

P.S. There is apparently an MT-32 emulator called MUNT. I just Googled "MT-32 emulator".

 
Posted : 04/06/2018 1:50 am
Angus
Posts: 8
Active Member
Topic starter
 

I guess I've got a lot to learn about GM. I played the game on the SoundBlaster 16 and it sounded alright. Maybe it would have sounded better on an MT-32, but doesn't the fact that it worked on an SB16 mean the game had to be able to produce GM? I don't think the SB16 could emulate MT-32.

I'm afraid I wasn't able to get very far w/Sekaiju. I couldn't find anything called List View. I'm running Sekaiju 5.3 on Wine 3.0.1.

It seems this MUNT thing requires DosBox. Does that mean I'd have to find a Dos app to play this MIDI?

 
Posted : 04/06/2018 8:18 pm
JohnG
Posts: 227
 

Hi Angus,

What is not generally appreciated is that the MIDI specification was initially created way back in, was it '83?
At that time there was no definition of what particular voice would be invoked by any Program Change message.
So Program Change #1 on one device might invoke Piano, but on another one might hear a saxophone, for example.
However, it might be that a manufacturer, in order to fulfil a specific market requirement, might make their set of voices very simialr to another device.
So an SB16 might have very similar voices to an MT32. I can't say for sure, I never checked it out although, way back, I started with an SB16 card (ISA version).

The later creation of the General MIDI (GM) standard changed all that by, amongst a few other things, creating a map of voices and what Program Change number invoked them.
GM didn't arrive until, as I recall, the early nineties. Maybe initially '91 with amendments in '94. --- I think.
If you read the Standards you'll see there's no mention of any mapping of voices to numbers in the original documents.
That only occurs in the separate GM standard specification ('91/'94), and a further mapping in the GM2 spec.('99?)
But, even today, MIDI devices don't have to be GM compatible. If they are they'll have a GM label visible somewhere.
But a pure synthesiser manufacturer, for instance, may decide not to try to emulate all the GM voices.

As GM, as a standard, was not created until some years later, no hardware that was created before this was exactly compatible, although a few may have come close.
It's for this reason that attempting to emulate old, pre-GM hardware, is fraught with pitfalls.

Especially so since both the SB16 and the MT32 used synthesis for sound creation as did most hardware at that time.
Holding samples in tiny amounts of memory or on floppies or diminutive hard disks was out of the question.
We are so spoilt today with our GigaBytes of memory and TerraBytes of disk space.
The first mainframe computer I worked on (a long, long time ago) had just 20 kilo-characters of memory. A character was six bits (plus parity), a word was 24 bits!

P.S. Further research shows me that many, including Creative, used the gaming card originally made by Ad-Lib, as a model for the voices.
It's quite posiible that Roland followed this voice layout too.

WRT Sekaiju, right click on the name of any track and you should see a pop-up menu.
In the pop-up you should be able to select "Show this track's Event List window".
Only one track's info will be shown initially, but you can select more tracks and what you want to see from the menu boxes on the RHS.

 
Posted : 05/06/2018 1:54 am
Angus
Posts: 8
Active Member
Topic starter
 

Hi John,

Looks like I got the dates of a few of these things mixed up. The bottom line is that the Wing Commander II came out before the SB16, so the latest SB card it could have supported was the Pro? In any case, I don't think the Pro is GM.

Figuring out what all those instruments are supposed to be seems like a lot of shots in a lot of dark--above my pay grade. If that's the explanation as to why the MIDI sounds fine until about 3:55 when it becomes a train wreck, then I don't see a way forward.

On the other hand, since WC2-45.MID sounds a lot like it should (on every platform I've played it), wouldn't it be easier to use that one as a guide?

 
Posted : 05/06/2018 7:20 am
Jason
Posts: 1
New Member
 

Hello,
I used to be able to play these files in Quicktime, but that app is defunct and I can't find any way to get them to play. Can anyone provide assistance?

Thanks,
J

 
Posted : 29/03/2019 5:53 am
Geoff
Posts: 1047
Noble Member
 

It would be a MASSIVE help if you could remove this query from this thread, and start a new thread. This query clearly has NOTHING to do with the the thread it is added to.

The two files attached are named as .MID, but looking inside them they are NOT MIDI/SMF files, although they MIGHT contain some midi-like data.

The files could be QUICKTIME audio format, as you say that QT would play them. I have a version of QT (I think) from about 1995 which is supposed to play QT videos, and MAY do something with QT Audio (but I'm guessing). I'll try to dig it out.

In any event, the files may be non-midi, and therefore outside the remit of this forum?

Geoff

 
Posted : 29/03/2019 2:21 pm
Lee T.
Posts: 6
Active Member
 

Angus,
I saw Rosegarden was used so wanted to look at this the same way.
I use a new Roland GO:KEYS 61k which has 16 channel enabled by default.

I see at bar 104 there is a switch from track 2 alone to track 2 plus track 5 then track 10, 12 , 19 and 21 are added at bar 108
but track 2 is not dropped.

This looks

 
Posted : 07/12/2019 11:32 am
Page 2 / 2
Share: