For people who are just getting started into vintage DOS gaming, MIDI music device support in games can be confusing:  What’s MIDI?  Why don’t MIDI devices play digital sound?  If all you have is MIDI, how do you hear sound effects?  This article will attempt to explain, as succinctly as possible, how MIDI devices became a part of DOS gaming music history, and how they were used for composing and playing back game music.  This article is not an extensive technical dive, but is meant to be an explanation for the novice DOS gamer who is new to the platform.

This information originally appeared as correspondence with Fabien Sanglard, who is working on the next book in his Game Engine Black Blook series.  His previous book, Game Engine Black Book: Wolfenstein 3D, is highly recommended.

What is MIDI?

MIDI is a synthesizer music control protocol, and stands for Musical Instrument Digital Interface. It was created in the 1980s as a way to connect keyboards, drum machines, other synthesizers, etc. together. Once connected, they could do things like have one instrument (like a keyboard) control another (like a drum machine), they could all refer to the same time sync signal, and also be connected to a computer using a computer MIDI interface. Once connected to a computer, the computer could record all of the notes passing along the MIDI protocol, effectively capturing the live performance.  Once captured, the computer allowed the user to edit the performance, make changes, and play it back out to all of the connected instruments.

MIDI instruments were connected to each other using cables employing a male 5-pin circular connector.

What was the Roland MT-32?
What was the Roland LAPC-I?

The Roland MT-32 was an external, standalone MIDI music synthesizer that you could connect with MIDI cables to a computer fitted with a MIDI adapter, such as the Roland MPU-401. This hardware allowed high-quality (for the time) music playback, better than the FM/OPL-based AdLib (and later Sound Blaster) cards. Sierra supported the MT-32 (and other sound and music devices) as early as 1987 and pushed for sound/music devices to be better-supported throughout the industry, leading with their own products first.  It is fair to credit Sierra with establishing higher-quality sound devices, and the MT-32 in particular, as a PC gaming “standard”.

The Roland LAPC-I combined the MPU-401 MIDI interface with an upgraded version of the MT-32 called the CM-32L into a single ISA card, so adding just that card would give you “MT-32” music playback, plus 33 additional sound effects.

The MT-32 was primarily a music synthesizer. Like any traditional musical keyboard synth, it had over 100 preset instruments, and some canned sound effects (gunshot, rain, explosion, wind, etc.). It also had the ability to change the parameters of those instruments and modify them to sound differently. Changing an instrument was done via SYSEX (SYStem EXclusive) MIDI commands. The MT-32 was not based on the Yamaha OPL series like the AdLib/Sound Blaster, but used a Roland-proprietary combination of prerecorded samples and subtractive synthesis.  You could not fundamentally create your own instrument or sound effect, but you could modify what already came with the MT-32.

The MT-32 only communicated via MIDI. MIDI does not have the ability to transmit PCM samples, so it had no ability to play custom digitized audio. The LAPC-I also lacked this capability:  While it was an add-in ISA card, it didn’t add digital sound playback to the existing CM-32L + MIDI interface combo it was designed to reproduce.  (A later card, the RAP-10, added this functionality.)

What was the Roland SC-55?
What was the Roland SCC-1?
Why were they a significant milestone of DOS gaming music?

The Roland SC-55 was the very first General MIDI standard device. The General MIDI standard was significant in that it defined 128 instruments that every device following the standard could adopt. Before General MIDI, there was no standardization: Instrument #4 could be “organ” or “flute” or “snare drum”, etc…  but in General MIDI, instrument #4 is always “Honky-Tonk Piano”. This standard was the real breakthrough for DOS gaming music, because it meant all MIDI music devices following the standard could reproduce the music nearly the same, as they would all be using the same instruments.  (That said, the SC-55 also had outstanding sound quality, which was also a breakthrough.)  Of the 128 instruments, 16 were sound effects.  An additional 46 percussive sounds (snare drum, hi-hat, etc.) were also available on a specific playback channel.

Like the MT-32, the SC-55 connected to the computer using MIDI cables and a MIDI interface installed in the computer. The Roland SCC-1 was like the LAPC-I: It combined an SC-55 and an MPU-401 onto a single ISA add-in card.

How was MIDI music for DOS games composed?

Standalone synthesizer modules like the MT-32 and the SC-55 can only be controlled via the MIDI protocol, sent over MIDI cables to the module. If you want to compose music to be played back on those modules, it is best practice to compose the music using the module itself. Because the module has no musical keyboard attached, it was common to use a synthesizer keyboard connected via MIDI to control the module. This was typically done by capturing a live performance on the keyboard into a MIDI note recording program called a MIDI sequencer, usually on a computer. Once captured on the computer, it could be edited like any media.

Why do I still need a Sound Blaster to hear sound in DOS games?

MIDI devices like the MT-32 and SC-55 were primarily designed as music synthesizer modules, and only contained a small number of sound effects.  The sound effects were so limited that most games didn’t bother trying to support them, so if one of those devices were all that you had, then you only heard music in a DOS game and no sound effects or speech. So, most DOS gamers also installed a card to play PCM sound, such as a Sound Blaster.  To hear both sound devices at the same time, users could run both to a mixer which output to a single set of speakers, but a cheaper option was to run the audio output of the MIDI module into the “line in” jack on a Sound Blaster, then connect the “line out” jack of the Sound Blaster to speakers.  Then the Sound Blaster internal mixer could be adjusted via software such that Line In was mixed into the output even while not recording.

While most games didn’t use MT-32 or SCC-1 canned sound effects, there were a few exceptions.  Sierra games using the SCI interpreter would occasionally use them (see Space Quest III for an example of MIDI sound effect use on the very first gameplay screen), and Another World (USA title “Out Of This World”) used the additional effects in the LAPC-I/CM-32L.

What music device should I pick when running a DOS game that has multiple options?

If you have an emulator configured for all possible DOS game sound standards — or, if you’re lucky and rich, a choice of sound hardware to put into a vintage DOS gaming system — then you should usually pick General MIDI with any sound extensions first, then General MIDI, then FM-based.  Here is a general list of sound options listed from best MIDI music quality to worst MIDI music quality; pick the highest option you have available:

1. General MIDI (Yamaha XG or XG extensions)
2. General MIDI (Roland SCC-1, Roland SC-55, Roland GS extensions)
3. General MIDI
4. Turtle Beach Multisound
5. Roland LAPC-1 or CM-32L
6. Roland MT-32
7. Sound Blaster 16, Sound Blaster Pro 2 (Yamaha OPL-4 or OPL-3 FM compatible)
8. Sound Blaster Pro (Yamaha OPL-2 stereo FM)
9. AdLib or Sound Blaster (Yamaha OPL-2 FM)
10. Creative Music System or Game Blaster
11. Tandy/PCjr 3-voice audio
12. PC speaker

Confusing expanded (EMS) memory with extended (XMS) memory happens all the time in the vintage PC/DOS computing hobby for those new to those terms.  Which is which?  What does my hardware support?  Do I need them?  This article will attempt to explain, as succinctly as possible, what each “type” of PC memory is and how it is used.  This article is not an extensive technical dive, but is meant to be an explanation of DOS memory conventions for the novice DOS user who is new to using DOS.

Protected Mode memory (DPMI, VCPI, DOS extenders, etc.) are not mentioned in this article because they typically don’t require any user intervention to work (ie. DOOM comes with DOS4GW.EXE and it just loads and works automatically).

Terminology used in this article

PC: Any IBM PC  or compatible system from the 1980s and 1990s running DOS.
8088: Shorthand for the entire 8086 CPU class, which includes the 8086, the 8088, the NEC V20, and the NEC V30.
640KB: The typical limit of memory DOS can access.  (Exceptions for DOS to access higher than 640KB exist, but are not mentioned to keep this article understandable for the layman.)

Quick history of memory on the PC

The 8088 CPU in the first IBM PC could access up to 1MB of memory.
Later, PCs built with the 80286 CPU could access up to 16MB of memory.
Later still, PCs built with the 80386 CPU could access up to 4GB of memory.

All PCs can directly access the first 1MB of memory.  Accessing memory beyond the first 1MB requires functionality only 80286 and later CPUs have.  Memory that extends past the first 1MB is called Extended Memory.

Because 8088 CPUs can’t access Extended Memory, special memory boards were created for them that fit into their system expansion slots.  These boards contain extra memory visible somewhere in the first 1MB, where the 8088 can see it.  The memory provided by these special expansion boards is called Expanded Memory.

Memory types in detail

EMS, XMS, conventional memory, UMBs… what does it all mean?  Here are some definitions:

Conventional Memory

What is it?  The first 1MB of memory visible to all 8088 and higher CPUs.  DOS is loaded into the area from 0-640KB and manages it.  The area above that, from 640KB to 1024KB (1MB), is typically used by the system hardware and not generally available to DOS programs.
How do DOS programs use it?  DOS provides function calls for allocating, resizing, and deallocating memory blocks for program use.  (Alternately, programs can directly access any portion of the first 1MB of memory they want to, although doing so can be risky if the programmer is inexperienced.)

Extended Memory

What is it?  Memory visible beyond the first 1MB, usually physically located directly on the motherboard, but sometimes provided by adding memory cards.
How do DOS programs use it?  Via an API called the eXtended Memory Specification, also called XMS.
What provides XMS?  XMS is provided by a driver called HIMEM.SYS, loaded by CONFIG.SYS at boot time.
How does XMS work?  XMS provides function calls that DOS programs can use to copy data between conventional memory locations and extended memory locations.

Expanded Memory

What is it?  Memory provided by special expansion cards.
How do DOS programs use it?  Via an API called the Expanded Memory Specification, also called EMS.
How is EMS provided?  Via a driver provided by the expansion card manufacturer.  Each card has a different driver and you must use the correct one for your card.  The driver configures the card for use, and provides EMS function calls that DOS or programs can use to access the memory on the card.
How does EMS work?  EMS boards provide a small “window” into the memory they contain, and this window is located somewhere in the first 1MB of RAM where 8088 CPUs can access it.  To access more memory, programs issue EMS function calls that “move” the window to a different area on the card, changing what portion of the larger memory shows up in the small window.

A quick note about EMS:  If you want to run a program that uses EMS, but you don’t have an EMS board in your computer, don’t fret: You can emulate EMS on any 80386 or higher.  This is done by loading a memory manager such as EMM386 or QEMM.  One of the services provided by a memory manager is to section off a portion of Extended Memory and present it in response to EMS function calls, like a real EMS board would.

Upper Memory

What is it?  The upper portion of Conventional Memory located between 640KB and 1024KB (1MB).  There is typically no user-accessible memory in this area, but 80286 and higher systems can relocate portions of Extended Memory into this area for use by DOS programs.
What is it used for?  The total amount of upper memory available is typically very small, between 64KB to 128KB, but this amount can still be useful for loading small memory-resident programs or drivers outside of Conventional Memory.
How do DOS programs use it?  XMS provides functions for allocating and deallocating memory in this area in units of “blocks”, called Upper Memory Blocks, or UMBs.
How are UMBs provided?  For 80386s and higher, DOS comes with HIMEM.SYS and EMM386 which, loaded together, provide UMB functionality.  For 80286s and lower, there are programs (such as QRAM or USEUMBS) that use “shadow ram” functionality provided by the 80286 system’s chipset to perform the mapping.
How does upper memory access work?  Once a program is provided an upper memory block via a UMB functional call, it can be accessed the same way conventional memory is accessed, ie. directly.

Guidelines to running DOS programs

1. First off, simply try to run a DOS program.  Most run without needing any special memory configuration.
2. If you have an 80286 or higher, always load HIMEM.SYS in your CONFIG.SYS file.  There’s no harm in installing it even if programs don’t use it.
3. If you want to load a single program on a 386 or higher to manage all of this for you, install QEMM.  QEMM will manage all of your memory and provide your programs with either XMS or emulated EMS based on what each program asks for.
4. Finally, if a program claims to support both EMS and XMS, choose EMS.  EMS is faster than XMS.