sneakernets / dmxopl Goto Github PK

To get myself back into things, I've decided to make a OPL/OPLL style General MIDI set. Nothing fancy here, just using the Sine Wave. I'm going to be working on DMXOPN2 again soon, so perhaps it's a form of "practice".

While I like the patch we have, it breaks in DMX due to DMX failing so hard with AM mode (Why??!)

Likely reverting to the old patch, but maybe I don't have to (Tyrian?)

Patch gave me a replacement today, I'll take a look at it.

FF6 and FF7 soundtrack exposed these issues:

• Make sure flute trill works in FF6 Prelude: Attack of some wind instruments are too low, rendering trills too quiet to hear in the mix. Other offenders include the Flute, Ocarina, and Piccolo patches.

Compared to SC-55 (or even that damned Microsoft GS Synth) they are totally different and hardly resemble what they're supposed to be, ESPECIALLY at lower octaves.

While I experimented with these two, I doubt I reached the optimum. Maybe I was going in the totally wrong direction ( ͡° ͜ʖ ͡°)

P.S. If you need more midis for testing purposed, I can "donate" some of mine. Downloaded them a goood while ago from an ancient japanese site, interestingly they're so good I listen to them UNIRONICALLY. Since japanese VN music from 90's was so damn complex, it's a good opportunity to test less used instruments. such as Tinkle Bell ;)

UPDATE: Msx box is done.

What I have is "okay" for now, but it just sounds like a trombone section sometimes. There may be no way to really solve this, but we can try!

Piano is still too loud for some midis and not loud enough for others. either I'm wrong or they are. HMM. Better figure this out.

Somewhat similar to #18 - rock organ patch is so "harsh" that it sounds like some kind of french horn or even tremolo strings ensemble (I even made an experiment swapping strings patch with rock "organ" one, it didn't sound that bad o.O)

There might be no way to solve this (at least I can't), but maybe...

P.S. If it helps a little - here is how I made rock organ in my OPN2 bank. Sounds rather hollow, though.

Doesn't sound good when used for guitar passages (funny that)

DMX Bugs require me to turn down the volume!

The project has been dormant for a while, being pretty much "done" for the most part, save for some drum kits and some GS/XG instruments that are being done "when I get around to it". But there are some instruments that I think need some work now that I've heard them enough:

• Pizzicato Strings: Sound too much like a steel drum somehow, needs some more timbre (but how?)
• Brass Section needs to have detune applied on attack, this once worked, but somehow got changed. This only affects a few midis.
• Shakuhachi needs a lot of work. It sounds pretty scratchy and bad in high ranges.
• Tubular Bells ADSR needs some cleaning up. Sometimes the bell gets cut off (ADLMIDI issue? who knows)
• Timbales need to be redone completely. There has to be a way to pull that sound off!
• Clean up toms on other drum kits.
• Finish the drum kits.
• Make Rhodes Piano a little better. This has been tweaked a lot since I started, and it may have gotten worse.
• Mayyybe tweak choirs a bit?
• Orchestra Hit needs redoing. It's too fuzzy. I had a better sounding one before. Find it maybe?
• Clarinet needs a little tweaking. It's much too "honk".

Hi,

I have some patches I'd like to contribute, but I am unsure of what process is in place. That is, how should individual patches be delivered to the repo?

Br.
Andreas

I cannot stay satisfied with the current Piano, even if it's a Yamaha preset. What I want to use aliases on Emulators. Now what?

DMX cuts the second voice of double voiced instruments when channel pressure is maxed out. Make sure that all double voices sound reasonably good with the second voice disabled.

Basically any double voices that use the extra channel to add more harmonics to the sound (but aren't 100% needed) need to be the second voice. It'd be embarrassing if the harmonics were the primary voice and the fundamentals are placed on the second voice.

As the title says. D_TENSE is way too loud.

Listening with a good Hi-Fi setup, the song here
https://instaud.io/private/0bea371c70161b688c67b509cdff7396aa9a0055

Has a lot of unneeded bass in the short guiro.

Since DMX isn't really a target anymore, Let's fix the cymbals. The hats are alright, it's the ride cymbals that are in trouble. I started work on the ride cymbal 1:

This sounds much better but it's still not "perfect" in matching the SC-55. Ugh.

Slap Bass patches need to be updated to a) be less loud, and b) match SC55/SC88 patches.

Ideas:

• Get Fourier graph of SC55 Slap Bass 1 and Slap Bass 2 sample (within usual range)
• Match patch output with sample on the first few harmonics
• Use a double voice to match even more harmonics (yet another double voice)

I had a discussion with one of the contributors, I may be going a little too ambitious on some instruments. Not sure what to do here. We're at this point, which I guess is good...

Whooops! I need to either revert a lot of stuff or just redo the volumes.

Finally found out what was breaking all my patches, you know, the ones that sounded like crap on DMX? Yeah... The carrier Level of the second voice is always the carrier level of the first voice. Let me explain.

When a Double Voice instrument is played, both voices' carrier waves will play as if they have the same volume level - which is determined by the first voice. So, if you have 63 for the first voice, and 45 for the second - well, too bad, you get 63 for both.

The reason I never ran into this is that I started only recently doing that kind of voice setup, while the others matched carrier wave volumes.

I don't know what other properties DMX reads from the first voice, but this is definitely one of them. This hasn't been documented anywhere, as far as I know. In fact, none of DMX has been documented, even though the source code is available.

See discussion at f54a1e9.
Instrument changes described, I might open a PR if needed.

A little trick I found with GS instruments (to knock it out of XG mode)

Line 989 of adlmidi_midiplay.cpp:

        bank = (uint16_t(Ch[channel].bank_msb) * 256) + uint16_t(Ch[channel].bank_lsb);

Should be

        bank = (uint16_t(Ch[channel].bank_msb)) + uint16_t(Ch[channel].bank_lsb);

This allows GS to work for now.

How do I use this with gzdoom? Maybe I'm missing something, but I haven't been able to find clear instructions anywhere.

I'm using gzdoom 3.1.0

D_DOOM intro is not playing properly, with very loud square waves playing in the bass. I think this might be an SC-55 sampling trick.

Thanks to a friend that noticed some of his midi files broke on some xylophone parts, sure enough, the top range that one would expect to play does not play. culprit? the one octave offset.

On some voices this is unavoidable. however, on those that can be fixed, let's fix it. Voices that are expected to play high ranges shouldn't be limited at all.

I may have to replace the toms with the older ones, as the new ones I just can't figure out how to make louder....

But I would like to salvage the idea, somehow, and make it work -- somehow.

Tone down the detune I guess? I hate that I have to do that...

Since I'm not targeting DMX exclusively anymore, I can finally have some hi-hats that aren't awful. If you feel like helping out let me know. they need to blend well with everything else, obviously.

Lee Jackson strikes again, abusing snares. Fix snares for now until libADLMIDI handles it.

Since this discovery today, Let's using the additional waveforms!

I'm pondering freezing this project and moving over to "DMXOPL3".

I've noticed that splitting up the XG/GS sets into distinct wopl files is a good way to start a massive headache. The only real problems between the two formats are related to the drumkits, so perhaps a software drum remapping (as BASSMIDI does) is the way to go here. Unfortunately, that's out of my hands.

The main issue is having GZDoom's LIBADL implementation follow these rules, as that is where my bank will have the most play - if GZDoom's LIBADL even sees GS/XG instruments at this point, which I don't think it does. I may be wrong though...

Since your latest update I'm having issues with the track ttd02.mid (Little Red Diesel)

Alot of notes are cut no matter how much OPL3 core I emulate.

Your can test with the latest version of my MIDI Player (1.6.9)

https://www.transporttycoon.net/music?recording=3

Sigh. DMX bugs came out of the woodwork when I made my new hats. Figure out a way to fix this.

Since Freedom's toms are so good, let's incorporate them somehow.

Cymbals need some cleaning up if possible. The current version is a compromise for DMX, the WOPL version wouldn't need to worry about this. Crashes could be better, ride bell too.

using my android player (libADLMIDI) the ttd02.mid track is rendering very weird (alot of bass note are cut)
it might be linked to the way gen_adldata compile your bank

you can find the track here
https://www.transporttycoon.net/music

One my friend is interested for SC8820 set as he using some instruments in his MIDI files on the SCVA thing.

I realize that this project has been dormant for over a year, mostly due to Real Life ™️ getting in the way. Now that Eternity Engine has ADLMIDI support and life has calmed down considerably, I can begin working on this project again.

A list of things I want to address:

• Credit - DMXOPL currently is a mishmash of contributions. Given the binary nature of the format we use, it's not really easy to track. I tried to keep tabs on this in my commits and release notes, but some uncredited stuff still slipped through.
• Bloat - There are a lot of files in the repo that are out of date and too large, including MP3 files. I'm not sure if there is a good way to remove these without breaking the repo, but these files will not be updated and may be removed in future.
• DMX support - This was dropped due to constant issues with exporting the WOPL file to the DMX format, which meant that several file formats had to be maintained separately. I want to investigate why these issues cropped up and hopefully fix them.
• Everything in #48 if possible.

Is the project still active?