clownmdemu - The Greatest Mega Drive Emulator Ever (Someday)

Discussion in 'Showroom' started by Clownacy, Jun 23, 2022.

  1. Devon

    Devon I'm a loser, baby, so why don't you kill me? Member

    Aug 26, 2013
    your mom
    In that case, I can at least give you some info about some of the CD reading BIOS commands that I have documented after digging through the BIOS a bit.
    • CDCSTART - Tells the CDC to start reading from the disc at the current sector. It queues up to 5 sectors
    • CDCSTARTP - Same as CDCSTART, except you can setup some flags for reading (see for more info)
    • CDCSETMODE - Just sets the read flags
    • CDCSTOP - Tells the CDC to stop reading data
    • CDCSTAT - Checks if the sector read queue is empty of not
    • CDCREAD - Tells the CDC to begin transferring 1 sector into memory
      • In manual transfer mode (MAIN CPU READ or SUB CPU READ), it gets sent to the host data register, where it is expected to be read word by word.
      • In DMA mode, it automatically transfers the data into the destination address (PRG-RAM, Word RAM, and PCM wave RAM can be set for copying into)
    • CDCTRN - If in SUB CPU READ mode, it handles copying from the host data register for you
    • CDCACK - Tells the CDC that the sector has been copied over into memory
    When implementing these, it's basically a must to at least look into how the CDD and CDC get data transferred, how header data is handled, and to know the different read modes so that you know how to account for them. The manual read modes involve reading the host data register one word at a time (of course, CDCTRN takes care of that for you for the Sub CPU side).

    The CD-ROM read BIOS commands (prefixed with "ROM") seems to just involve seeking to the designated sector you want for reading from, where it then automatically calls CDCSTART, and also pausing and unpausing reading. The CD drive also cannot play music and read data at the same time, it can only do one or the other, so keep that in mind.

    Hope this helps!
  2. Clownacy

    Clownacy Retired Staff lolololo Member

    Aug 15, 2014
    That's great: I was worried that I'd have to rely solely on the Mega CD BIOS Manual and the macros in the DDK's 'SEGADTS\32X\DEMO\CD_V2A\INC\CDBIOS.INC' to figure this stuff out, so the insight helps!
  3. Clownacy

    Clownacy Retired Staff lolololo Member

    Aug 15, 2014
    [Blog post]

    If you've been keeping up with the development of my Mega Drive emulator, you'll know that I've recently been trying to add support for the console's addon - the Mega CD. Last time, I was able to get a cartridge-based Sonic ROM-hack to boot. This hack made fairly-minimal use of the Mega CD, making it easy to at least get to boot. Now, I've reached the next milestone by getting an actual commercial Mega CD game to boot! That game, of course, being Sonic CD!

    The most immediate hurdle to getting a proper Mega CD game to boot is the CD itself: my emulator was designed around reading data from a cartridge dump, so I had to add an entire new mechanism for fetching data from a CD image. While a cartridge is 'flat mapped' into the 68000's memory, allowing the CPU to see and access all of its data at once (like an array), a CD must have its data copied into a RAM buffer before the 68000 can read it. The CD is far larger than the Mega CD's RAM (700MiB vs 768KiB), so only small portions of the disc's data can be loaded at once. Data is read from the disc by making various calls to the console's BIOS.

    To test my emulator's ability to emulate said BIOS calls and correctly read data from a CD dump, I created my own Mega CD port of my Visual Sound Test homebrew. In the process of making this port, I learned about the Mega CD's start-up process, as well as which combination of BIOS calls is required to load data from the CD. In a day or two, I had a port that would run in Genesis Plus GX (another emulator that supports the Mega CD), and I began expanding my emulator to be able to run it.

    This led to the second obstacle which needed to be overcome: the Mega CD's boot process.

    The Mega Drive's boot process is incredibly simple: in the first 0x100 bytes of the ROM is a 'vector table', which is a standard data structure that the 68000 uses to initialise itself: it initialises the stack pointer and tells the 68000 where to begin executing code. Once the 68000 begins executing code, the game's code can go about initialising the rest of the Mega Drive hardware and begin playing the game.

    The Mega CD's boot process is far more complicated: there is a small header at the very start of the CD's data which details the location and size of two blobs of code: the Initial Program (IP) and System Program (SP). The IP is loaded into the Mega Drive's WORK-RAM, to be executed by the Mega Drive's CPU, while the SP is loaded into the Mega CD's PRG-RAM, to be executed by the Mega CD's CPU. The SP has a header that contains offsets to various call-backs: one for initialisation, one for the 'main' function, and one for the vertical interrupt (V-Int).

    The way the CD header works is oddly complicated, having fields that either do nothing or do not function in the way that you would expect. The data from this header is read by the Mega CD's BIOS, which it then uses to load the CD sectors that contain the IP and SP blobs into their respective parts of memory. The SP header is parsed in order to install the proper call-backs, which the BIOS then proceeds to call. The IP and SP programs both run, taking over from the BIOS and running the game. It's extremely unlikely that an entire game can be fit in the IP and SP blobs, so instead these often act as a 'chain-loader', communicating with the BIOS to load more sectors from the CD which contain more code and data. Sonic CD in particular has files on its CD that are called 'IPX' and 'SPX', which contain code that is responsible for loading and running the actual game.

    To handle all of this complicated initialisation logic, the Mega CD relies on its BIOS. However, my emulator does not use the Mega CD's BIOS, instead favouring the 'High-Level Emulation' approach. Because of this, my emulator can't simply run the BIOS - it has to be the BIOS, doing all of its tasks in its absence. This is comparable to playing Sonic 1, not by emulating the Mega Drive, but by just making your own Sonic 1 from scratch. Because of this, my emulator has to manually perform the entire Mega CD boot process itself, reading the first CD sector, parsing the header, loading the IP and SP programs, parsing the SP's header, and finally running them.

    My Visual Sound Test port's IP will colour the screen red to show that it is running. If the SP is also running, then the IP will detect it and colour the screen green. After that, the SP will load the Visual Sound Test as one big binary blob into WORD-RAM, pass WORD-RAM to the MAIN-CPU, and IP will jump into it, completing the initialisation process and running the Visual Sound Test.

    With proper implementation of the boot process, the IP and SP were working together to colour the screen green. One thing was stopping the Visual Sound Test from running: the lack of BIOS calls.

    As said before, games load data from the CD by calling the BIOS. The BIOS is given a value that denotes an operation to be performed, and the BIOS does it. Without an actual BIOS, my emulator has to fill its role, receiving these values and performing the task that the BIOS would have done in response. This is fairly simple to do: to run a BIOS call, the SUB-CPU jumps to address 0x5F22, causing the 68000 to read a word of machine code from address 0x5F22. By detecting reads from address 0x5F22 in my 68000 emulator's bus event call-back, I can perform the requested BIOS task and then return the machine code of an 'rts' (return) instruction to the 68000, causing it to exit the BIOS call and resume the game code.

    By analysing Sega's official BIOS documentation, sample code provided with the so-called 'DDK' (an SDK from the 90s), and disassemblies of Sonic CD and the Mega CD BIOS, I was able to implement rough approximations of several BIOS calls related to reading the CD. With this done, my Visual Sound Test homebrew was able to boot and run perfectly!

    (This is an old screenshot, so just pretend that this is running in my emulator and that it's the Mega CD port.)

    There was one small issue that's worth mentioning here: Mega CD games are required to insert some "security code" (a Sega logo splash screen) at the start of their IP. For whatever reason, this code would crash in my emulator, so I made my emulator simply skip it for now. This code is not essential in any way, to my knowledge.

    With my ported homebrew booting, I moved onto finally getting Sonic CD to boot! There was much debugging, as various mistakes in my emulator were causing the boot process to fail very early on, but eventually files were finally being properly chain-loaded: from IP to IPX to MDINIT to BURAMINIT... which would indefinitely hang.

    BURAM (Back-Up RAM) is where save data is stored. BURAM has its own series of related BIOS calls which needed implementing (notably, they use address 0x5F16 instead of 0x5F22), but upon stubbing these out (all returning an error code), I was greeted by this:


    Finally, a sign of life from Sonic CD! I don't know a lick of Japanese, but I assume that this is an error screen of some sort. It won't let you proceed past it, so I made some of the BURAM BIOS calls pretend to run successfully to bypass it. Unfortunately, the game would freeze between this stage of initialisation and running the title screen. This was hard to debug, and had me stumped for a few days. Confused and demotivated, I forced myself to spend a few hours today trying to find the problem, only to discover that it was the result of an interrupt handler in my stub SUB-CPU BIOS not preserving the registers! This caused a loop that was intended to clear a data buffer to begin clearing the code buffer, including itself. Upon fixing this, I ran the emulator one more time and was blown away to see this:


    Victory! Sonic CD boots! Unfortunately, it crashes when I try to start a new game, but still, this is a great milestone!

    It's a strange thought that one of my emulator's first milestones was reaching the title screen of Sonic 1, and now, after all this time, the latest milestone is reaching the title screen of Sonic CD. Things sure have come a long way from this emulator being a 68000 interpreter and VDP renderer held together by a little bus logic. Soon, I hope to have Sonic running around Palmtree Panic Zone!
    Last edited: Oct 11, 2023
    ProjectFM, Hame, Pacca and 2 others like this.
  4. Clownacy

    Clownacy Retired Staff lolololo Member

    Aug 15, 2014
    In the last post, I mentioned that my emulator does not use the Mega CD's BIOS program because it favours the 'High-Level Emulation' approach. In this post, I'll go over what that means and how it will affect my emulator.

    Other emulators, such as Genesis Plus GX and Kega Fusion, require that the user provide their own copy of the Mega CD's BIOS, which has to be extracted from a Mega CD. Without this, the emulator will be unable to play Mega CD games. I cannot bundle this BIOS with my emulator because of copyright limitations (Sega owns the BIOS and I don't have the legal right to distribute it). However, expecting the user to provide their own copy of the BIOS is not an acceptable solution to me, as it creates busywork for the user: good software should be as easy to use as possible, and expecting the user to buy a Mega CD, extract its BIOS, and then configure the emulator to use it, is the furthest thing from easy. Likewise, expecting the user to download the BIOS from some random website is also unnecessary busywork (though, legal or not, a person certainly has a moral right to download software that they already own a physical copy of).

    This means that the most user-friendly option is to forgo requiring the original BIOS entirely. This leaves me with two options: make my own open-source BIOS, or make my emulator not need a BIOS at all.

    Making my own BIOS should not require much explanation: I just have to create my own BIOS that implements the same API. If I am able to do so without copying any copyrighted material (code, graphics, or sound) from the original BIOS, then my BIOS will legally be my property, and I can freely distribute it with my emulator.

    Making my emulator not require a BIOS at all is a bit more complicated: typically, an emulator recreates the behaviour of a console's hardware, and then runs the console's software on top of it. This is known as Low-Level Emulation (LLE), because the 'lower layer' (the hardware) is the part that is being recreated by the emulator. However, it is possible for an emulator to instead recreate the 'higher layer' (the software), performing what is called High-Level Emulation (HLE). This is how an emulator can be made to not require a BIOS: instead of recreating the behaviour of the underlying hardware, and then running the original BIOS, which interacts with said hardware, the emulator can instead recreate the behaviour of the BIOS itself, making the original BIOS (as well as its underlying hardware) redundant.

    There are various trade-offs between using the original BIOS, making my own BIOS, and making my emulator not need a BIOS:

    Using the original BIOS is the most accurate solution, as it is the same software that is used by a real Mega CD. This means that all of the behaviours and bugs will be the same. However, the original BIOS is burdened by copyright issues, making it the most awkward option for the user. In addition, it is an awkward solution for me, the emulator developer, as I have to make my emulator emulate parts of the Mega CD hardware that are used by the BIOS. This includes the CDC, CDD, and BuRAM. The final downside of this approach is that the BIOS remains a 'black box' - obscure, obfuscated software. This makes it difficult for the user to understand what the BIOS is doing and how to make modifications to it. If a person is given software, then they deserve to be able to understand and modify it as easily as is feasible, and relying on a binary blob completely undermines that.

    Making my own BIOS rids the user of the difficulty of obtaining a copy of the original BIOS, but places a burden on me by requiring that I program an entire BIOS from scratch. The new BIOS may be inaccurate to the original one, differing in intended behaviour, suffering from bugs that the original did not, and missing bugs that the original had. For the new BIOS to act as a true replacement for the original, it should provide all of the same features and interact with the same underlying hardware. Because of the latter requirement, I would still have to make my emulator emulate the CDC, CDD, and BuRAM hardware, among others. On the upside, my new BIOS would be open-source, fully documented, and deobfuscated, allowing the user to understand, modify, and even distribute it with relative ease. However, I say only 'relative' ease because the BIOS would be at least partly written in Motorola 68000 assembly language, which is a fairly unknown language in today's world. The rest of the emulator is written in ANSI C, so it may be too much to expect the average person browsing the emulator's source code to comprehend the BIOS's code. One last upside of creating my own BIOS is that it could be used outside of my emulator: it could be used with other emulators, or even with a real Mega CD!

    Making my emulator not need a BIOS at all also rids the user of the burden of sourcing a copy of the original BIOS, and has the same downside of having questionable accuracy. However, it bypasses the need for my emulator to emulate the CDC, CDD, and BuRAM hardware, as it is (at least mostly) encapsulated by the BIOS. No BIOS; no hardware. This is similar to how WINE doesn't need to emulate a whole PC just to run a Windows program. I think some of the CDC's registers are interacted with directly by games' SUB-CPU programs, but that shouldn't require emulating the entire CDC. Code related to replacing the functionality of the BIOS would be part of the emulator itself, and therefore would be written in the same language as the rest of the emulator - ANSI C. That would make this the best option when it comes to users being able to read and modify the code. However, naturally, this code is tied to my emulator and cannot be used with other emulators nor a real Mega CD. To some extent, there may be a performance upside to this approach, as my emulator would be running native code instead of running emulated code in a CPU interpreter and emulating a variety of additional hardware components. HLE code is also significantly easier to write than a custom BIOS, as HLE code does not need to interact with the underlying hardware that a true BIOS would, making me to free to implement functionality as I see fit.

    I hope that this comparison has made it clear why I have opted for High-Level Emulation.

    Performing Low-Level Emulation of either an original or custom BIOS is straightforward enough: just emulate the underlying hardware as with any other part of the Mega Drive and Mega CD hardware and then run the BIOS just like it were a game. While it is not simple in execution, it is simple in concept. High-Level Emulation is more complicated, due to it crossing the line between software and hardware. So, how exactly does one make a Mega CD emulator not require a Mega CD BIOS?

    First, let's go over the boot process: using a real BIOS, this would be very simple: just boot the BIOS by resetting the Mega Drive's 68000 CPU with the BIOS's vector table mapped to address 0x000000, and the BIOS will handle communicating with the CDC and other hardware to load and execute the Initial Program (IP) and System Program (SP) from the CD. With HLE, however, the emulator must do this manually: when an attempt is made to reset the Mega Drive's 68000 CPU, the emulator must intercept this and load the IP and SP on its own, configure the Mega Drive and Mega CD to the expected state, and then allow the 68000 CPU to resume execution.

    The BIOS is also responsible for interrupt handling, however it appears to mostly just bounce the interrupts to a secondary interrupt table in RAM, which can be modified by the game in order for it to install its own interrupt handlers. This can be recreated by simply populating the vector table and secondary table in RAM with the expected values on boot.

    Next is the BIOS call process: when a SUB-CPU program wants the BIOS to do something, it stores a value into data register 0 and then jumps to address 0x5F22. In the original BIOS, the code at 0x5F22 is the start of a function which uses the value in data register 0 to select an action, which it then performs. These actions include lighting-up the LEDs on the front of the Mega CD, opening and closing the disc drive, playing music from the CD, and reading data from the CD. To recreate this with High-Level Emulation, I have to detect when the SUB-CPU has jumped to address 0x5F22. The way I do this is by exploiting the emulated 68000 CPU's bus mechanism. As a real 68000 CPU runs, it is constantly fetching code through 'bus events', which involves the CPU essentially pausing itself, signalling that it wants some data from an address, signalling that address, and then waiting for it to be given that data. This is very similar to a function call, and that is exactly what my emulator uses to emulate bus events. In the function, my emulator will check if the address and the CPU's program counter (the address of the code currently being executed) are both 0x5F22, and if so, it performs the action that corresponds to the value inside data register 0. The function will then return the machine code for an 'rts' instruction, causing the emulated 68000 CPU to immediately return to the code that it came from, as if it had ran a proper BIOS function.

    Under-the-hood, my emulator is free to perform its actions however it wants, regardless of how the original BIOS did it, so long as it provides the same API. This allows me to make certain quality-of-life improvements that would not be possible if the emulator were using the original BIOS. For instance, Genesis Plus GX stores all of its Mega CD save data in a single file, which contains the contents of BuRAM. Because my emulator does not emulate BuRAM, however, it is free to deviate from what a real Mega CD does: notably, because the BIOS's BuRAM API is file-oriented, I can make each game's save data its own file on the user's PC!

    And that's everything I can think of to say about High-Level Emulation: its upsides, its downsides, how it differs from Low-Level Emulation, and how it is done. I hope that this has been interesting. I realised that I really skimmed over this in my last post, which is unfortunate because the nitty-gritty technical stuff is my favourite part of the Dolphin, Citra, and Yuzu progress reports, and it's what I hoped to provide by starting a blog in the first place.
    ProjectFM, JGamer2151, Pacca and 2 others like this.
  5. JGamer2151

    JGamer2151 Well-Known Member/Lurker Member

    Dec 1, 2020
    Very interesting topic about emulating the Sega CD BIOS via HLE; I know that to me there are some emulators out there that can emulate the BIOS of various consoles via HLE such as the Connectix Virtual Game Station (which was a PS1 emulator for the Mac/PC), and others.

    Now while it is not this project’s purpose to implement this (and I certainly don’t, considering with the current state of the emulator that you have), but I suspect that the BIOSes for the 32X will also have to be emulated via HLE in order to get around the aforementioned legal issues with distributing BIOSes in emulators (which is a big no-no within the emulating community), especially when it comes to your emulator.
  6. Devon

    Devon I'm a loser, baby, so why don't you kill me? Member

    Aug 26, 2013
    your mom
    Implementing the 32X boot ROMs actually wouldn't be nearly as difficult.

    The 68000 side is basically just a 256 byte header that points to the secondary vector table on the cartridge whenever there's a soft reset that still leaves the 32X mode enabled (long story short, in 32X mode, the cartridge mapping changes and no longer sits at the beginning of the memory map, so in the original spot, that header lies so that the soft reset can work). The header also appears even when the RV flag is set (temporarily reverts the cartridge mapping back to Genesis mode, away from the 32X stuff to safely do DMAs from ROM and other stuff), since the 32X mode is still enabled.

    This 256 byte header, though, contains 2 functions located in the dummy vector table entries that basically safely deal with the Sega mapper (for bank switching) or SRAM for you.

    The SH-2 side is just 2 boot ROMs (1 per CPU) that sets up the CPUs, perform a security and checksum check, and load the SH-2 program from the cartridge into the 32X's SDRAM. If in Sega CD mode 2, it waits for the program to be copied into the 32X's frame buffer (it waits for the Genesis to send "_CD_" via communication register 0 before doing the transfer to SDRAM).

    If you wanna see in detail how the Sega CD 32X setup is done, here's this.

    I also have some quick disassemblies of the 32X boot ROMs. The use of .lits are just guesses, but the code is there.
    Last edited: Oct 16, 2023
    JGamer2151 likes this.
  7. NayTheGamer

    NayTheGamer rawr..... Member

    Sep 30, 2022
    I would like to point out that this emulator doesn’t support my Rom hack, it only gets far as the SSRG splash screen but after that, it crashes. This isn’t the game as it works fine on every emulator (exodus not included) and even works on Real hardware, I hope this gets fixed once the New version is out aka V3.0
  8. Clownacy

    Clownacy Retired Staff lolololo Member

    Aug 15, 2014
    [Crosspost from the blog]

    I was hoping to release this update after I finished adding support for the Mega CD, but progress on that has ground to a halt. Instead, the focus of this update will be the frontend! From features to refactoring to an entire port, here's what's new in clownmdemu!

    Frame Advance
    Frame Advance is a fairly mundane feature for typical users but a very handy tool for developers: it allows for advancing the game by only a single frame, which can be useful for examining a bug or verifying that animations occur correctly. It is activated by pressing the fast-forward hotkey while the emulator is paused, and can be combined with the rewind key to advance to the previous frame instead of the next.

    At first, I figured that this would be an awkward feature to add, for code-flow reasons. However, the opposite turned out to be true: it couldn't have been more convenient to add!

    Debug Logging Window
    Previously, the frontend would output debug information to the terminal (known to you Windows users as the command prompt), but when emulation goes catastrophically wrong, so much debug information is output to the terminal that it causes the entire emulator to slow to a crawl and becomes unresponsive. To avoid this, debug information is now output to a Dear ImGui window. By default, logging is disabled, minimising the performance impact of the emulator issuing debug information in the background. Even when enabled, however, performance is still far better than before, with instances where the window is flooded with debug information still resulting in a responsive emulator. If, for whatever reason, information absolutely must be output to the console, then an option is provided to reenable it.

    There is one downside to this: the logging window is very RAM hungry. However, to avoid crashes, the logging window will detect when RAM has ran out and clear the log, freeing much of the RAM and allowing the emulator to continue running normally.

    In a Mega Drive cartridge or Mega CD disc, code is stored as machine code, which is essentially just a bunch of hexadecimal numbers.

    This is Sonic 1's boot-code in machine code form:
    This is very hard for a person to read, so the more practical way of viewing code is in assembly form. The process of converting machine code into assembly is called 'disassembly'. My emulator's frontend is now capable of automatically performing disassembly, allowing the code of the emulated game to be viewed with ease!

    This is Sonic 1's boot-code in assembly form:
    The 68000 disassembler is actually an off-shoot of my CPU interpreter library - clown68000. This is because, long ago, I made an effort to split the interpreter into two parts: the interpreter itself, and a kind of database library that describes the functionality of the 68000. The idea behind this was to minimise the work that would have to be redundantly repeated every time the interpreter was rewritten, or a new 68000-related tool was made. This plan ended up paying off, as the database library was not only used to automatically generate part of the interpreter, but now components of it are used by the disassembler. In particular, the instruction and operand decoding were made trivial by the library, allowing a quick-and-dirty prototype to be made in just one day. Being closely related to each other, the two now share a Git repository.

    While this will surely be useful for the developers of ROM-hacks and homebrew, I'm hoping to use this for reverse-engineering games that don't work correctly in my emulator. In particular, I'm hoping that this will be the key to getting Sonic CD working.

    Frontend Refactor
    Originally, the frontend was written in ANSI C, but it became C++98 when Dear ImGui was integrated, as that was the language that it (and, more importantly, its API) were written in. Later, Dear ImGui switched to C++11 and the frontend switched with it. Throughout all of this, however, the frontend never made much use of C++'s features, remaining very C-style. This has finally changed, as now the frontend has been converted to proper C++: 'NULL' has been replaced with 'nullptr', references are used instead of pointers, casts are now C++-style, and many components has been split off into their own classes. The migration to modern C++ features should improve the frontend's code hygiene, such as through the more-explicit nature of C++-style casting.

    The frontend has also been refactored to split much of the code from 'main.cpp' to separate files. 'main.cpp' was getting excessively large, which was making it hard to navigate, so I began splitting whatever I could to other files. This was done long ago with the debug menus, but poor encapsulation made them a bit of a mess. In this new wave, the audio stream was split to its own self-contained class, as was the window subsystem, the emulator wrapper, and the file dialogs.

    68000 Interpreter Refactor
    It is possible for an emulator to be excessively accurate: an emulator could recreate the exact behaviour of a console on the transistor level, but the game running on the emulator would behave exactly the same as it would on an emulator that recreates the console's behaviour on a logic-gate level. The only difference is that one emulator is far more performance-intensive than the other. There comes a point where additional accuracy does not benefit emulation, as the external behaviour is already perfectly accurate, even if the internal behaviour is not.

    This same issue applies to CPU emulation. You may have heard the term 'cycle-accurate' before; this refers to the accuracy of the emulator's timing: on a given cycle (a fixed time-step that things like CPUs operate on), a cycle-accurate emulator's state (RAM, CPU registers, etc.) will be the exact same as the original hardware on the same cycle. Neither of the CPU emulators in my Mega Drive emulator are cycle-accurate: the closest one is my Z80 emulator, which is accurate only during the cycles on which an instruction ends. This is because instructions are executed over the course of multiple cycles on a real Z80, while my emulator completes every instruction instantly in a single cycle, and simply does nothing for the remaining cycles. This may sound like an unimportant technical detail, but this does affect when certain operations (such as accesses to the VDP's ports) occur. This creates a difference in behaviour that could cause the emulated game to behave differently. This is imperfect emulation.

    To correct this, I could rework my CPU emulators to be 'truly' cycle-accurate. That is, on each cycle, the emulator does the exact same tasks as a real CPU. However, like described earlier, this would be overkill: many of the tasks done per cycle are strictly internal to the CPU, and are completely invisible to anything outside of it. Since a game exists outside of the CPU running it, this means that these internal behaviours are completely irrelevant to a game running exactly the same as it does on a real console. One example of such an invisible task is adding two numbers together using the CPU's registers. In contrast, a task that would not be invisible is sending the result of that addition to RAM, since the RAM exists outside of the CPU.

    Since cycle-accuracy is too accurate, and being accurate only on the cycles that instructions end is not accurate enough, I have settled on the perfect in-between: being accurate only on the cycles that a bus event occurs on. A bus event is a transfer of data between the CPU and the rest of the system, making it the fundamental external task of a CPU. Hopefully you can tell where I am going with this: by making bus event behaviour accurate, I make all external behaviour accurate, and therefore make the behaviour of software running on the CPU emulator 100% identical to a real console (at least when putting aside other emulated components of the console that can influence code execution such as the bus and VDP).

    With that massive preamble out of the way, I have been refactoring the 68000 emulator in preparation for making it accurate on a bus-event level. This involved splitting the instructions into multiple steps, so that they can be completed over the course of multiple bus events instead of all at once as they did before. Each step is made into a unique function, then pointers to these functions are grouped into lists, one per instruction, and then, when an instruction needs to be ran, it is done by calling each function in the instruction's list, with bus events occurring between each function call.

    In hindsight, by doing all of this, I have implemented microcode into my 68000 emulator. While the emulator is still not yet accurate on a bus event level, all of the preparatory refactoring is now out of the way, leaving me with far less work to do to finally make the switch.

    Improved High-DPI Support
    On Windows, the emulator's frontend already enjoys high-DPI support, making its fonts as crisp as possible! I assumed that the code for this would work on other platforms too, but I heard from a Mac user that it was horribly broken!

    What it should have looked like:

    What it looked like:

    Without access to a Mac, and no understanding of what could possibly be going wrong, I was unable to fix this. Around the same time, I noticed that the DPI scaling was overkill on Linux, causing the font to be far larger than it needed to be. With all of these issues in mind, I disabled high-DPI on any platform that wasn't Windows. This left Mac and Linux with blurry fonts (and blurry rendering in general), which I wasn't happy with.

    High-DPI is one of the SDL2 library's shortcomings: while its support for the feature on Linux and Mac is good, Windows support was entirely missing for the longest time. More recently, it gained some support for Windows, but it had to be activated in a different manner to the other platforms and came with a bunch of caveats and edge-cases, defeating the whole point of SDL2 being a platform abstraction library.

    I wanted to try again to implement high-DPI support on Linux, so I enabled the dormant original high-DPI code to see what would go wrong. When the emulator window appeared, I was surprised to see that it exhibited the exact same bugs as on macOS. As it turns out, on Linux (with Wayland, at least), SDL2 uses the same high-DPI mechanism as it does on macOS. This meant that, if I could fix high-DPI on Linux, then high-DPI on macOS would be fixed too!

    At first, I tried supporting high-DPI the same way that Dear ImGui's SDL2 example does, but it was terrible: all it does is force SDL2 to convert the rendering coordinates from pixel coordinates to 'device coordinates', to match the mouse movement events and window size, which are also measured in device coordinates. Device coordinates remain the same across DPIs, whereas pixel coordinates get smaller and smaller, which is why using device coordinates universally is one way to support high-DPI. The problem with making Dear ImGui render using device coordinates manifests when using a DPI that is not a multiple of 100%, such as 150%: in this situation, the font and other native-resolution graphics will not be aligned with the screen's pixels as they are rendered, causing them to appear blurry, and also aliasing is introduced to the clip rectangle logic, causing letters to occasionally be cut off at an edge. The end result is extremely ugly rendering.

    (Left: Good Rendering - Right: Garbage Rendering)

    Appalled by this, I instead tried the opposite approach: rather than make the coordinates consistent by converting the renderer to device coordinates, I converted the mouse movement events and window size to pixel coordinates. This did result in awkward aliasing in mouse positions because SDL2 brilliantly only exposes mouse coordinates as integers instead of floating-point numbers, but otherwise this works perfectly. This does create more work for me though, as all of the calculations for rendering the frontend must be upscaled in accordance with the DPI because a higher DPI means more pixels. However, all of the work for this was already done when high-DPI support was added for the Windows port, since everything was already measured in pixel coordinates on that platform. So, in the end, all I did was make Linux and macOS behave like Windows, and everything magically worked! Gee, if only there was a library called SDL2 that did that... that would be great...

    Web Browser Port
    One of the main features of my emulator is its portability: the core emulator is written in portable ANSI C, the frontend is written in C++11 and only dependant on cross-platform libraries such as SDL2 and Dear ImGui, and the build system is CMake. In addition, while the build script supports linking system libraries, it is able to build these libraries manually if they are absent. All of this combined made the emulator almost immediately compatible with Emscripten!

    For those that do not know, Emscripten is a toolchain that allows C and C++ to be compiled to either JavaScript or WebAssembly, instead of the usual CPU-specific assembly. JavaScript and WebAssembly can be embedded into a website, allowing it to be ran in a web browser. Emscripten also exposes various APIs to C and C++, allowing code to interact with the web browser to do things like render graphics and receive keyboard and mouse input. SDL2, which serves as a platform abstraction layer, supports Emscripten's APIs, granting software that is built upon it compatibility with Emscripten.

    The 'main' Function
    There is, however, one incompatibility between typical C/C++ code and Emscripten: the 'main' function. Normally, the 'main' function only returns when the program closes. For a constantly-running program like an emulator, this means that there needs to be an infinite loop inside the function. But, with Emscripten, infinite loops are not allowed. Because of this, software that targets Emscripten must instead use the function to register a callback, which Emscripten will then run on the function's behalf a specified number of times per second. This callback can be used to iterate the program in the same way that the function's infinite loop would have.

    Because of this incompatibility, I did need to refactor my emulator's frontend slightly. It resulted in some nice decoupling, however: now, the emulator frontend and the 'main' function are entirely separate, with the frontend occupying its own source file and exposing an interface that allows for initialising, updating, and deinitialising the frontend. The 'main' function is the sole occupant of the 'main.cpp' source file, calling the frontend's iteration function in an infinite loop in standard builds, and registering it as a callback in the Emscripten build.

    File IO
    Another one of SDL2's shortcomings as a platform abstraction library is that it does not provide any way to create a file dialog to allow the user to select a file for the program to load or save. Because of this, my emulator's frontend manually implements these file dialogs for each platform, using the Win32 API for Windows, and Zenity and kdialog on Linux and the BSDs. Since Emscripten is compatible with none of these, it requires its own platform-specific code.

    I was able to find a handy MIT-licensed single-header-file library that implements Emscripten file dialogs, however, its API was much higher-level than what the frontend was designed for. The reason for this is that, due to running in a web browser, software suffers from strict limitations, with one such limitation being that files on the user's computer cannot be directly accessed. Instead, software must ask the browser to access the file on its behalf.

    Because of this, the library does not simply return the path of the file that the user selected: instead, for loading a file, the library makes the browser read the whole file into a memory buffer and then returns a pointer to it, and, for saving a file, the library takes a pointer to a memory buffer and makes the browser write its contents to the file.

    While, previously, the tasks of reading and writing data to and from a file were the responsibility of the frontend, they now had to become the responsibility of a library. To achieve this, file IO had to be abstracted to operate purely on sending and receiving buffers instead of file paths. This abstraction layer would be implemented as a C++ class - the 'FileUtilities' class - which acts as a wrapper around the 'emscripten-browser-file' library in Emscripten builds, and standard file IO in non-Emscripten builds.

    The Result
    With all of this work complete, my browser is able to load and run games within the confines of a web browser!

    What is great about this is that I can link to it directly from my blog's Projects page, allowing anyone that is interested in the emulator to try it out! It would be nice to do this with some of my other software too, such as ClownMapEd. With some work, I could even make this auto-load my various ROM-hacks, allowing them to be tried-out in the browser too!

    If you're interested in trying this port out, you can find it at

    Vastly-Improved Audio Playback
    I have never been happy with the frontend's audio delivery system: due to emulation being synchronous, the frontend is not simply able to render audio on command, instead needing to wait for audio to be generated by the emulator before it is able to pass it on to the operating system to be played through the user's speakers. In order for there to not be gaps in the audio, audio must be generated as fast as it is played. However, the audio-generation process is extremely complicated (involving generating the audio output of two different sound chips - FM and PSG - at two entirely different sample rates - 53kHz and 224kHz - and then resampling them both to a common sample rate - typically 48kHz - and then mixing them together), meaning that audio is never exactly the correct size. Because of this, it's possible for audio to be generated slightly faster than it is played. When this happens, a gradual build-up of audio occurs, making the audio more and more delayed. For the longest time, this was worked-around by detecting when too much audio was being generated and simply throwing away some of the excess. This kept the audio from ever exceeding a certain amount and becoming latent, but also caused the audio to audibly 'skip', producing irritating popping and crackling sounds.

    This problem came to a head when adding support for 50Hz (PAL emulation) to the Emscripten port: the popping was so annoying that I simply had to do something about it. I decided that I would make the frontend resample (stretch or squash) its audio in accordance with how much audio build-up there was. This way, there would be no skips in the audio, making playback as smooth as could be. I spent the next few days experimenting with different methods of making the audio playback speed self-adjusting. It was a demoralising few days, as everything I tried failed in one way or another: either the self-adjustment would be too weak and it would fail to respond to fluctuations in framerate, causing audio skips or latency, or it would be too aggressive, causing the audio to constantly warble as the pitch was greatly increased, only to be greatly decreased to prevent too much audio being generated, only to be greatly increased again to prevent too little audio being generated, and so on and so forth.

    On the verge of giving up, I remembered that somewhere in RetroArch's documentation was a paper that described a method of 'dynamic rate control' that was just what I needed. I found it, implemented its formula, and was happy to hear it work perfectly! The frontend now adjusts the speed of its audio to always maintain a certain amount of generated audio at all times, preventing both skipping and latency. One of the RetroArch formula's features is that the speed adjustment is kept to a minimum, preventing the change in pitch from being audible. At last, the frontend's audio system is something that I can be proud of!

    Unicode File Path Support on Windows
    To end things off, here is a feature that I implemented just now! The frontend no longer has any problem with accessing files and folders whose names contain non-ANSI characters. This means that if you have a file whose name is written in Japanese in a folder whose name contains emojis, then the frontend will be able to load it! Previously this would just fail, as the internal way of encoding strings was incompatible.

    Unfortunately, the font system doesn't support the entirety of Unicode, so many characters will appear as question marks when viewed in the frontend itself, like in the 'Recent Software' list.

    So here it is - one big update! How typical of me to want to implement Mega CD support, only to end up doing a whole bunch of other things instead. Hopefully these new additions will be enough to tide people over until then!

    Source code:
    Windows executable:
    Try it in your web browser:
    Last edited: Dec 25, 2023
    Pacca, JGamer2151, Kurk and 3 others like this.
  9. RobiWanKenobi

    RobiWanKenobi Python Developer and ASM enthusiast Member

    Sep 10, 2022
    United States
    Something you should know, Sonic The Hedgehog 3/3&K's Data Select Menu doesn't show up right in this emulator.
    It looks like:

    It should look like:
  10. Clownacy

    Clownacy Retired Staff lolololo Member

    Aug 15, 2014
    Try it in your web browser:
    Windows EXE:
    Source code:

    Even-Further-Improved Audio Playback
    It was pointed out to me that my Mega Drive emulator has another source of audio distortion: while v0.5 had managed to fix the audio skipping that was caused by audio being generated too quickly, there was still crackling that was caused by the FM or PSG audio being truncated by 1-3 samples every frame.

    The reason for this is that the audio resampling process is prone to rounding error, meaning that if 400 samples were desired, 401 samples may be produced instead. Because there are two resamplers running simultaneously (one for the FM audio and one for the PSG audio), it's possible for each of them to output a different number of samples. When this happens, the number of samples which are output is capped to the smaller of the two. This causes the audio with the larger number of samples to skip slightly, creating a crackling noise.

    At first, I tried fixing this by making the recently-added dynamic audio sizing system apply to the FM and PSG audio separately, but, while this did eliminate the crackling noise, it caused the two audio sources to frequently desynchronise with each other, causing music to play with disjointed timing.

    A proper fix was achieved by reworking the resamplers to always generate an exact number of samples. By doing this, the FM and PSG audio are always perfectly synchronised, and no excess audio frames are produced. At last, perfect audio delivery!

    Frontend Debugging
    While working on v0.5's dynamic audio sizing system, I became interested in monitoring the audio buffer. I had added some code to print statistics to the debug log, but this was messy and so it was never included in the released version. Even once the system was complete, I wanted to be able to keep on eye on how it was operating to ensure that it was working correctly. Instead of keeping the hacked-together debug logging code around forever, I made it into a proper debugging menu!


    Various useful little bits of information are displayed here, particularly the target and average numbers of audio frames. The closer these numbers are; the better: if the average becomes too low, then audio drop-out occurs, and if the average becomes too high, then latency occurs. Additionally, there are statistics for the framebuffer upscaling, which is nice for ensuring that the non-integer scaling is working optimally. The list of SDL2 drivers has been moved here from the About menu too, making it easy to see which APIs SDL2 is taking use of. Here you can see my painfully contrarian OpenGLES2, Wayland, and PipeWire configuration. Because OpenGL 2.1, X11, and PulseAudio are so last year.

    C++ Refactoring
    The frontend has been receiving more C++ refactoring to make it less C-like and more exception-safe. Particularly, instead of using 'fake' constructors and destructors that are just regular methods, the frontend's classes have been converted to use real constructors and destructors. This had many implications for field lifetime and initialisation, requiring even further refactoring. I've also eliminated the need for manually-defined destructors from a number of classes, leveraging smart pointers and the like to make the automatically-generated ones sufficient.

    The strictness of object-oriented code has helped expose a number of design deficiencies already, such as the configuration loader calling methods of the audio output object before it had even been initialised. It has also encouraged better factoring of the code by nudging me towards splitting the audio device logic from the audio queuing system, which coincidentally makes it simpler to port to other APIs. Overall, I've enjoyed getting to know modern C++ better, and think that it benefits the codebase quite a bit.

    Hopefully, this shake-up to the codebase has not introduced any bugs nor instability. If it has, expect to see some small hotfix updates in the coming days.
  11. Clownacy

    Clownacy Retired Staff lolololo Member

    Aug 15, 2014
    [Cross-post from the blog]

    Recently, I fixed a couple of inaccuracies, and I thought they would be interesting to read about:

    Fixed the Title Cards in Sonic Crackers
    I've tested many Sonic games and ROM-hacks in my emulator, but I've never tried the famous Chaotix prototype - Sonic Crackers. If I did, I would have noticed an obvious bug that affects its title cards:

    (It may be hard to believe, but this is not what it's supposed to look like).

    This bug had flown under the radar for over two years until it was helpfully pointed out to me in a GitHub Issue. I found this bug very interesting because it appeared to stem from a mature and battle-tested part of the VDP emulator, so I was surprised that there was still an inaccuracy within it.

    Upon investigation, I discovered that the bug in my emulator was being triggered by a bug in Sonic Crackers! The tiles for the title card are uploaded using DMA transfers, however the game's DMA transfer routine is bugged: immediately after performing the DMA transfer, the routine attempts to reupload the first word of tile data (this is presumably to work around a known bug in early Mega Drives that causes the first word of a DMA transfer to be corrupt), however, it accidentally only writes half of the address command. This is important, as, when half of an address command is written to the VDP, a 'latch' flag is set to signal that the next word written will be the command's other half. The DMA transfer routine will then loop and begin to prepare another DMA transfer, starting by writing a register command for the DMA length. Because the latch flag is set, however, this register command is instead interpreted as the second half of an address command, completely breaking the DMA transfer, causing the bug that is seen above.

    This is not the case on a real Mega Drive: there, the register command is processed correctly, allowing the DMA transfer to successfully occur. This shows that the latch flag only applies to words that are not register commands, whereas my emulator was making it apply to all words, register command or otherwise.

    But that is not the only issue that this bug exposes: later in the routine, an address command is written to the VDP in order to trigger the DMA transfer. With the latch flag still set, the emulator would receive the first half of the command, notice that the latch flag is set, and use it as the previous, incomplete command's second half. As with before, this breaks the DMA transfer.

    Once again, this is not the case on a real Mega Drive. The most likely reason being that the latch flag is cleared whenever the VDP receives any word, even if it isn't part of an address command, preventing the DMA transfer from being broken by the incomplete address command.

    Addressing both of these inaccuracies finally allows the title card to work properly:

    (It may be hard to believe, but this is what it's supposed to look like).

    This issue illustrates why accuracy is so important to emulators: while well-behaved code will obediently follow the rules to the letter, bugs will invoke all kinds of edge-cases and undocumented behaviour that no sane code ever would. In this case, Sonic Crackers' bugged code relied on quirks of the VDP's address command latching behaviour that would never be tested in normal circumstances.

    Fixed the Data Select Background in Sonic 3
    A longstanding bug has finally been fixed! Ever since Sonic 3 first booted in this emulator, the background of the Data Select menu has looked like this:


    With the help of the emulator's debugging menus, I could see that the background's rows had been split into two groups:


    This struck me as very, very odd, as there is no way that this could ever work on a real Mega Drive: what's happening is that the game has set the VDP to render in 128x32 mode, but it has loaded the background in 64x32 mode. There is no bug here: this is what the game actually does, and the VDP is doing exactly what it should in response.

    So why is it that this works on a real console? The foreground needs to be in 128x32 mode to fit its graphics, and it's not possible for Plane A and Plane B to be different sizes. So how could Plane A possibly be correctly displaying 64x32 data in 128x32 mode?!

    That is a trick question - did you figure it out? Here's the answer: the background is not displayed by Plane A - it's displayed by the Window Plane. Yep, that obscure feature that's never used by Sonic 1 and 2 is used for a menu background of all things in Sonic 3.

    But what's special about the Window Plane? Well, one of its "features" is that, unlike Plane A, it cannot scroll. Because of this, being 128x32 would make no sense for the Window Plane, as most of it would be forever off-screen. As a result, the Window Plane is hardcoded to either 64x32 and 32x32, depending on what resolution the game is rendering at. I never knew about this hidden gimmick, so when I added Window Plane emulation almost a year ago, I made it match the size of Plane A and Plane B.

    Fixing this makes the background display correctly at long last:

    Last edited: Feb 1, 2024
  12. Clownacy

    Clownacy Retired Staff lolololo Member

    Aug 15, 2014
    (Cross-post from the blog).

    Here is a small update that fixes a few inaccuracies in order to get more games working.

    Try it in your web browser:

    Fixes for Sonic 3's Data Select and Sonic Crackers' Title Cards
    I have already covered this in another post, but for those not in the loop, the VDP emulation logic contained a few inaccuracies that caused visual corruption in a couple of Sonic games. These issues have now been addressed, allowing these games to work as they do on real Mega Drives.

    [​IMG] [​IMG]
    [​IMG] [​IMG]

    Fix for Thunder Force IV's Gradual Corruption
    A user, Pinkerton, recently pointed out a very strange bug: the game Thunder Force IV would start off working fine, but gradually corrupt over time.

    [​IMG] [​IMG]

    This is unlike any bug that I have ever seen before in this emulator: games usually either work or they break completely - they do not slowly corrupt little by little whilst otherwise remaining completely stable.

    An invaluable tip that Pinkerton provided was that the debug log kept mentioning the use of the 'STOP' instruction, which is a 68000 CPU instruction that was not yet emulated. I noticed that the Sega splash screen was playing at too fast a speed, so it was apparent that the screen was using the STOP instruction for controlling its framerate. The rest of the game played at the correct speed, and the debug log stopping complaining about the STOP instruction after this screen was over, so it was clear that only the Sega screen used the STOP instruction. Still, I figured it would be worth implementing the STOP instruction to get the screen working properly, then afterwards I could focus on the graphical corruption.

    The STOP instruction is not used by the main Sonic games, so I am not very familiar with it, hence why it was not emulated. Upon researching it, I found that the instruction essentially causes the 68000 to cease processing instructions until an interrupt occurs. Given that there is an interrupt - Vertical Interrupt - that occurs once per frame, this instruction becomes a useful way of idling the CPU until the next frame. It would be unfortunate if no interrupts were set to occur, as the CPU would end up waiting forever, so the STOP instruction is also able to set which interrupts are enabled and disabled.

    Implementing the STOP instruction only took a couple of minutes and a few lines of code, and, with that done, the Sega screen was running at the correct speed. Surprisingly, this also fixed the gradual graphical corruption in the game itself!

    But how? The main game does not even use the STOP instruction! My guess is that the game relies on the STOP instructions in the Sega screen to set the interrupts to a sane configuration, and that, without it, the game would receive interrupts whilst performing sensitive operations like communicating with the VDP, causing said communication to be corrupted and leading to the graphics being broken.

    Regardless, I am glad to have fixed the bug. Thunder Force IV was a game that I was meaning to try out sometime, so it's nice that I can play it in my emulator.

    Get Combat Cars, Micro Machines, Dynamite Heady, and Gunstar Heroes to Boot
    For as long as my emulator has existed, it was unable to boot Combat Cars and Micro Machines, even after massive improvements in accuracy to both the 68000 and Z80 interpreters. This has always stumped me.

    After informing me of the bug in Thunder Force IV, Pinkerton also noted that two Treasure titles - Dynamite Heady and Gunstar Heroes - failed to boot. With the new 68000 disassembler, I set about diagnosing the issue.

    I was surprised to see that the games were stuck reading from the Z80 bus request register. To my knowledge, reading that register will provide a flag that indicates whether the bus request is in-progress or not. The Sonic games frequently wait for this flag to be clear. However, these Treasure games were different: they were waiting for the flag to be set instead. Because my emulator does not work like an actual Mega Drive, bus requests complete instantaneously, so the flag is never set. Because the flag is never set, the games wait indefinitely, resulting in their failure to boot.

    Upon looking into the details of the bus request register, I discovered that I was wrong about what reading the register does: the flag provided does not indicate whether the request is in-progress, but rather whether the bus is held or released. What this means is that, if you're releasing the bus, you wait for the flag to be set, and, if you're requesting the bus, you wait for it to be clear.

    With this inaccuracy corrected, Dynamite Heady and Gunstar Heroes now boot!

    [​IMG] [​IMG]

    Not only that, but I later noticed that Combat Cars and Micro Machines finally boot too!

    [​IMG] [​IMG]

    According to Pinkerton, this change gets Alien Soldier to boot as well!

    Bonus: Struggling with Windows XP Compatibility
    Going back to at least the second release of this emulator, I have tried to maintain compatibility with Windows XP. Why? Because I can, and because it was pretty convenient: the frontend's main dependency, SDL2, supports Windows XP, and Visual Studio provides a (deprecated) toolchain for targeting Windows XP. I also figured that maintaining XP support was a good way of benchmarking the frontend's portability: if it doesn't run on Windows XP, then what are the odds of it running on niche homebrew platforms?

    Recently, I set up a Windows XP virtual machine just to test this emulator, so I could enjoy the satisfaction of seeing it running. I double-clicked the EXE and... it failed. No error messages at all; it just failed.

    I did some research, and found that Visual Studio broke its Windows XP compatibility after version 16.4. This affects the runtime library, which I normally link statically. Switching to dynamic linkage and installing an older runtime library on Windows XP should bypass this issue, so I did just that. The emulator still crashed.

    At this point, I considered Visual Studio to be a dead-end, and decided to switch to an open-source toolchain instead. MSYS2 would have been a good option, but it not only dropped support for Windows XP, but also Windows 7, so that would not work here.

    That left one more option: Arch Linux's MinGW-w64. Years ago, I used it to produce Windows XP-compatible builds of the Sonic 1 and 2 remasters. I rebooted into Linux, installed the appropriate packages, cross-compiled the emulator, copied the binary over to my virtual machine, double-clicked it, and... it failed.

    This time was slightly different: the error message mentioned a missing DLL import, that being the function GetTickCount64, which was added in Windows Vista. This was a known issue with MinGW-w64's libwinpthread. A recent commit from the end of 2023 should have fixed this, but it appears that this has not yet been released in an update. Fortunately, an Arch User Repository package is available for compiling the latest libwinpthread sources. With that done, I built a new binary and ran it. It still crashed.

    Well, damn. Now what? At this point I was well and truly defeated, and gave up. Windows XP support with a modern toolchain is infeasible in 2024... or so I thought. Just this morning an idea occurred to me: what if SDL2 was failing to initialise because hardware acceleration was not available? VirtualBox no longer provides hardware acceleration to Windows XP guests, and SDL2 tries to create a hardware accelerated renderer before falling-back on its software renderer, so perhaps the emulator would work if I just forced SDL2 to skip hardware acceleration. To do this, I had to set the 'SDL_RENDER_DRIVER' environment variable to 'Software'. After that was done, I ran the EXE and...


    It works!

    Unfortunately, this is not as convenient as compiling with Visual Studio on Windows, but I think that it is acceptable given that I already have to switch to Linux to compile the Emscripten port. This may be the very first release of the emulator whose pre-built EXE actually works on Windows XP! Though, that does not mean that the EXE will run badly on modern versions of Windows either: it will still support high-DPI displays and leverage newer Windows APIs like Direct3D 11 and WASAPI! That is the magic of SDL2, dynamically bridging the gap between the program and the operating system's underlying APIs!
    Last edited: Feb 16, 2024
    JGamer2151, ProjectFM and Hame like this.
  13. Clownacy

    Clownacy Retired Staff lolololo Member

    Aug 15, 2014
    (Cross-post from the blog)

    Try it in your web browser:

    Recently, I have been informed of various games that do not work correctly in my emulator by BlastBrothers, who has opened several issues on the emulator's GitHub repository.

    Properly Initialised Emulator State
    One of these issues mentioned missing audio in Battletoads: upon booting, the title screen would not play any music. Despite this, music would play when starting a level, and the title would play music if the game is reset. These quirks suggested that this was an matter of the emulated Mega Drive having invalid state upon boot.

    After some investigation, I found that the Z80's bank was not being initialised at all. This caused it to default to the value of the memory that it occupied, which, on x86 Windows at least, is 0xCD. Because the Z80 bank was set to such an unusual value, the game's sound driver was reading music data from the wrong location, causing no music to play.

    It is worth mentioning that this would not have been an issue were the emulator written in C++ rather than C, as class field variables are always properly initialised, be it explicitly or implicitly. Using C means that I have to do all of this manually; there is no safety net here.

    Vastly-Improved VDP Command Behaviour
    Another game that was misbehaving was The Adventures of Batman & Robin, which had missing and corrupt graphics in its opening:

    [​IMG] [​IMG]

    Additionally, there were various Electronic Arts games whose EA splash screen did not appear correctly:

    (This looks pretty cool in motion).

    What these both had in common was the debug log being flooded with warnings about invalid access modes and the VDP being written to while it was in read mode. It appeared that communications with the VDP were going horribly wrong. This struck me as very odd, considering that I had recently made VDP communication far more accurate in order to get Sonic Crackers working correctly.

    To debug this, I consulted Nemesis's VDP test suite (yes, the same Nemesis that the compression format is named after). His test suite is a ROM that you run on a Mega Drive emulator to verify the accuracy of various behaviours. Safe to say, my emulator does not pass many tests:


    What was interesting was that the emulator did not pass the 'partial command port writes' test. I found this strange because I thought my emulator did support those: partial commands are cached and later combined with their missing halves when they are sent by the 68000. Is that not what a real VDP does?

    As it turns out, no, it is not. As this incredibly useful document by Genesis Plus developer Charles MacDonald explains, the VDP actually applies partial commands immediately, and does not wait for their missing halves to arrive. This means that any software which relied on this behaviour was operating on outdated VDP state, as the commands that were being uploaded were not doing anything. I also found that my previous understanding of command latching was completely incorrect, and only worked in Sonic Crackers by complete accident.

    After addressing these inaccuracies, the two games began looking much better!

    [​IMG] [​IMG] [​IMG]

    This massive overhaul also incidentally fixed the stray pixels in Mega Man: The Wily Wars. It has been a long time coming!

    [​IMG] [​IMG]

    Improved Accuracy of DMA Fill

    Despite all of these improvements, Thunder Force IV continued to display garbage in its opening:


    MacDonald's document helpfully pointed out that this game required accurate DMA Fill emulation. DMA Fill is a feature that allows VRAM to be filled with a repeated value (though undocumented, CRAM and VSRAM can be filled as well). I thought that I had already implemented this accurately, yet, once again, it turns out that I had not. Oddly, when writing each byte, a real VDP does so to the current address XOR'd with 1. I do not understand why it does this. Regardless, my emulator failed to reproduce this behaviour, apparently causing the above bug. After applying a simple fix, the bug disappeared!


    Made 68000 Byte Writes More Accurate

    There was one small problem with the previous fix - it broke Wily Wars:


    But how could this be? Even Nemesis's test ROM confirmed that this was the correct way to emulate DMA Fill! After over an hour of debugging, my mind was not changed: my emulator was absolutely working correctly. So then why was this bug happening?

    As a reminder: Wily Wars uses DMA Fill to create several blank tiles in VRAM. This is what they normally look like:


    The problem is that the third and fourth pixels of each tile are not being set by the DMA Fill, causing a 'hole' to appear in them.

    To explain why this could not be a problem with the DMA Fill itself, I shall go over how a DMA Fill occurs:

    The final step of activating a DMA Fill is a standard memory write: this causes a word of data to be written to the destination memory at the current index, after which the index is incremented. Following this, the VDP enters a loop of writing the upper 8 bits of the word to the destination memory at the current index XOR'd with 1, until its counter reaches 0.

    Let us illustrate this more directly; here is what happens when a DMA Fill of 0xFF is performed at index 0 with an increment of 1 and a length of 3:
    • Standard memory write occurs, writing 0xFF00 to address 0.
      • 0xFF is written to address 0 (0 XOR 0).
      • 0x00 is written to address 1 (0 XOR 1).
      • Index is incremented to 1.
    • Byte-writing loop begins:
      • 0xFF is written to address 0 (1 XOR 1).
      • index is incremented to 2.
      • 0xFF is written to address 3 (2 XOR 1).
      • index is incremented to 3.
      • 0xFF is written to address 2 (3 XOR 1).
    The result of all of this is that the four bytes starting at address 0 are 'FF 00 FF FF'. Since a tile is 4-bits-per-pixel, this explains why the third and fourth pixels of each one is missing. The DMA Fill is working exactly as it should, so how does Wily Wars rewrite the skipped pixels, and why does it not work in my emulator?

    To find this out, I used my emulator's handy disassembler to view the code in the game that was triggering the DMA fill. Here it is:

    0006CAF8: movea.l 4(sp),a0
    0006CAFC: lea     ($C00004).l,a1
    0006CB02: move.w  sr,-(sp)
    0006CB04: ori.w   #$700,sr
    0006CB08: move.w  #$8F01,(a1)
    0006CB0C: move.w  ($FFBE8C).l,d0
    0006CB12: bset.l  #4,d0
    0006CB16: move.w  d0,(a1)
    0006CB18: move.l  $12(a0),(a1)
    0006CB1C: move.w  #$9780,(a1)
    0006CB20: move.l  $1C(a0),(a1)
    0006CB24: move.b  7(a0),-4(a1)
    0006CB2A: move.w  (a1),d0
    0006CB2C: btst.l  #1,d0
    0006CB30: bne.s   $6CB2A
    0006CB32: move.w  ($FFBE8C).l,(a1)
    0006CB38: move.w  #$8F02,(a1)
    0006CB3C: move.w  (sp)+,sr
    0006CB3E: rts

    What is most important is the 'move.b 7(a0),-4(a1)' instruction. This is the instruction which specifies the value to fill the RAM and ultimately triggers the DMA Fill. Normally, this instruction writes a word ('move.w'), but this one moves only a byte ('move.b'). As soon as I saw this, I knew exactly what the problem was.

    While reading posts on the SpritesMind forum, I learned of a quirk of the 68000 that my emulator did not recreate: when performing byte-sized memory writes, the 68000 will repeat the byte in the top and bottom 8 bits of its external data bus. Since the VDP only supports word-sized port accesses, it will read both of these bytes together as a single word. This means that Wily Wars' code, rather than writing 0xFF00 to the VDP, is writing 0xFFFF.

    Following the above steps again, but using 0xFFFF instead of 0xFF00, produces 'FF FF FF FF' rather than 'FF 00 FF FF'. This is why Wily Wars does not have holes in its tiles on real Mega Drives.

    Correcting this fixed Wily Wars once more!


    Implemented DMA Copy

    There is not much to say about this one: I do not know of a single game that uses DMA Copy, but Nemesis's VDP test suite includes a few tests for it, so I figured that I would finally implement the feature.

    DMA Copy copies bytes from one place in VRAM to another. It shares DMA Fill's quirk of XOR-ing the destination address with 1, but DMA Copy also XORs its source address. Its other quirk is that it always increments its source address by 1, not by what the destination address increment register has been set to.

    I was able to make this share most of its code with DMA Transfer, so the two might share circuitry in a real VDP.

    Made Debug Log and Disassembler Output Copyable

    This is a small usability improvement to the frontend: the Debug Log and 68000 Disassembler windows now allow the user to select their text. In particular, this is for copying the text and pasting it elsewhere. For instance, the user could disassemble a function and then copy it into a text editor for further analysis or even to use it in their own project. Likewise, the debug log can be copied for pasting into a GitHub issue to report a bug.

    Moved Configuration Files to Standard Location

    The frontend has two configuration files, and, up until now, they have always been located in the 'working directory', which is usually where the executable itself is. If the emulator is ran via the command line, however, then the working directory will be whichever directory the command line is currently viewing. This can lead to configuration files cluttering various user directories without warning.

    To solve this problem, configuration files are now stored in a standard per-user location. On Windows, this will be the 'AppData' folder, while, on Linux, it will be '.local/share'. As well as avoiding clutter, this has the added benefit of making all copies of the emulator share a single set of configuraion files. No more changing the keyboard bindings for the umpteenth time just because the EXE had been moved to another folder.

    Improved Accuracy of Sprite Masking and Overflow

    The Mega Drive has a limit to how many sprites it can render at once: there is a limit on how many sprites can be drawn across the whole screen, and there is a limit to how many sprites and pixels can be drawn on a single horizontal line. When this limit is reached, sprites simply stop being drawn. This can lead to graphics disappearing or flickering during heavy sprite usage, such as when defeating a boss in Sonic Spinball. One place that this can easily be seen is in Sonic the Hedgehog's Spring Yard Zone boss battle:


    Here, the blocks along the bottom of the screen combined with the lives counter result in the pixel limit being reached, causing parts of the blocks to vanish.

    The game's title screen makes use of this to achieve a particular effect: numerous junk sprites are placed off-screen in order to reach the sprite limit, making the bottom of Sonic's sprite disappear. If it were possible to see all of the sprites that the game is rendering, including the off-screen ones, here is what it would look like:


    Were the sprite limit not emulated, the title screen would look very odd:


    However, there is another way to hide parts of sprites like this: if a sprite has an X coordinate of 0 (which is very far off-screen to the left), then no more sprites will be rendered on the lines that it occupies. This type of sprite is known as a 'sprite mask'. Sonic the Hedgehog 2 made use of this feature for its title screen, allowing it to achieve the same effect as its precursor without the need for so many junk sprites.


    Sprite masking and the sprite/pixel limits have many edge-cases and little-known behaviours, meaning that many emulators do not emulate them accurately. To help make this a thing of the past, Nemesis created yet another small but exhaustive test suite.

    I had actually used this to improve my emulator long ago, however there was a certain test that always failed:


    This one had to do with a particularly odd edge-case within another edge-case: a sprite mask will not prevent sprites from being drawn if no other sprites have drawn before it unless the previous line reached its pixel limit. While this is simple enough to comprehend, I did not want to have to write a mess of spaghetti code to recreate the behaviour, so I left it unaddressed until I could come up with an elegant way to implement it. Fortunately, I finally came up with such a solution, and now all tests finally pass:


    On another note - you might be wondering what I was using to see those offscreen sprites earlier. Well...

    Added a Sprite Debugger

    At long last, two years after adding debuggers for everything else, I have finally made a sprite debugger! In fact, I made two:


    As seen in this image, one debugger lists each sprite in detail, while the other shows how the sprites are composed into the final image. Notably, sprites can be drawn far off-screen, and this debugger allows them to be seen! Additionally, with this, the Mega Drive's rendering can be dissected down to each and every layer!


    Wrapping Up

    And that is everything of note in this update! Not since v0.4 has this emulator seen such a slew of fixes and features! Like with v0.4, development may slow down for a while after this, as writing such a big blog post and working on one project so much can cause quite a bit of burn-out. Hopefully this update is stable and user-friendly so that there is no huge need for another update any time soon.
    Last edited: Apr 1, 2024
  14. Devon

    Devon I'm a loser, baby, so why don't you kill me? Member

    Aug 26, 2013
    your mom
    What good timing! I was actually recently learning more and more about how the 68000 and VDP work. This will definitely be useful to read over.
    I had learned about this and was even talking to someone about this just yesterday even.
  15. Devon

    Devon I'm a loser, baby, so why don't you kill me? Member

    Aug 26, 2013
    your mom
    Heya... I've been helping make some magic happen on the Mega CD side~

    peachy, JGamer2151, ProjectFM and 3 others like this.
  16. Red2010 is now

    Red2010 is now Active Member Member

    Apr 24, 2023
    Somewhere in Spain
    Someday support for 32x games will arrive. I can't live well without a V.R Deluxe race every so often
  17. SSRGteen117

    SSRGteen117 i like to s1, s1 Member

    Jan 6, 2021
    Angel Island Zone
    looks very cool! cant wait to try it out someday.
  18. Clownacy

    Clownacy Retired Staff lolololo Member

    Aug 15, 2014
    The workaround that I was using to keep the images in these posts small has stopped working for some reason, so, if you want to see the post as it was meant to be, read it at my blog.


    Try it in your web browser:

    It is finally here! With my job resuming tomorrow, I decided that I should release what has been completed so far before I have far less spare time. The main feature of this update is greatly-improved Mega CD support, allowing Sonic CD to be played from beginning to end. But that is not all: there are also bugfixes, improvements to accuracy and user-friendliness, and the emulation of even more Mega Drive features!

    Fix After Burner II's Audio
    After Burner II is fascinating: it plays PCM samples using the PSG channels, rather than the DAC channel. Because there is only one DAC channel, but multiple PSG channels, this allows the game to play multiple PCM samples at once, which is something that is not possible to do otherwise without mixing samples on one of the CPUs!

    However, the game was failing to output any PCM samples in my emulator. Using the debugger, it could be seen that the PSG was indeed being used to play PCM samples, but it just was not audible.

    Fortunately, I had a suspicion about what the problem could be: the documentation that I had used when creating my PSG emulator mentioned that, at a low-enough frequency countdown setting, the square wave ceases to alternate. I found this to be questionable, so I did not implement it in my emulator. However, the PSG would be outputting such a high frequency at this setting that it would be not only imperceivable to human hearing, but filtered-out entirely by the audio resampler. So, I did as the documentation said, and stopped the wave from alternating. And, with that, After Burner II's samples became audible!

    Fix Jim Power's Flipped Graphics
    While investigating another inaccuracy, I stumbled upon this bizarre bug:

    After having made so many improvements to VDP communication lately, I was baffled to see yet another problem with it crop-up. It appeared that each longword of data that was being written to the VDP's plane maps had its words reversed. After some testing, I concluded that this was not a problem with DMA transfers nor VRAM byte-swapping, so I began tracing longword-sized writes from the 68000 to the VDP's ports. Upon detecting one, I used the emulator's disassembler to view the code that performed it, and this is what I found:

    00004260: bsr.s   $4268
    00004262: swap    d3
    00004264: tas.b   d3
    00004266: move.l  d2,d1
    00004268: swap    d3
    0000426A: move.l  usp,a6
    0000426C: move.l  d3,(a6)
    0000426E: move.l  d1,-(a6)
    00004270: rts     

    Upon seeing this, I knew exactly what the problem was: the game was using a pre-decrement longword write, and it expected each word to be written in reverse order. In other words, it expects the latter word to be written to the VDP first, and then the former word afterwards.

    While it was a bit of a pain to make my 68000 interpreter recreate this behaviour, I soon had it done, and I could enjoy Jim Power's hypnotic parallax scrolling bug-free!

    Seriously, try moving to the right in this game - it gets weird.

    Use Software's Name as the Window Title
    Unlike other emulator developers, I see my emulator as less of a Mega Drive simulator and more of a Mega Drive compatibility layer. With that in mind, the frontend should provide as 'thin' a layer between the running software and the user as possible. One area where this was not the case was the window title: if the user was playing Sonic the Hedgehog, then the window's title should be 'Sonic the Hedgehog', but, instead, it would always be 'clownmdemu'. That issue has now been fixed.

    How does the emulator know what the software's name is? Sega required that software made for the Mega Drive have a header, and this header includes a name. Here is the header for Sonic the Hedgehog:

    The header includes many things: the target console, a copyright notice, the software's unique serial code, revision number, a checksum, compatible accessories, ROM size, RAM usage, regions, and two names. One name is for Japan, and the other is for everywhere else. The emulator will display the name that matches the region that the emulator has been set to.

    Notably, the names are encoded in SHIFT-JIS rather than ASCII, allowing them to contain Japanese!

    Columns' header seemingly contains gibberish...

    ...until you decode it as SHIFT-JIS!

    This feature was slightly awkward to implement: while the names are SHIFT-JIS, the window title must be UTF-8, so the names have to be converted. SDL2 has built-in text encoding conversion, but it does not support SHIFT-JIS, so the emulator has to perform this conversion on its own. Fortunately, I had already tackled SHIFT-JIS-to-Unicode conversion before for the CSE2 project, so I could just copy the code which I had written for that. That code would output UTF-32, however, so I needed to write an additional converter to turn UTF-32 into UTF-8.

    With all that done, users can now see the name of the game that they are playing!

    Implement the V Part of the H/V Counter
    I dread implementing the H/V counter, as it is a hardware feature which I am very unfamiliar with, which does not seem to be particularly well-documented, and which would require significant refactoring in order to be emulated 'properly'.

    The H/V counter is a value that the 68000 CPU can read from the VDP to tell which pair of pixels (or a 'slot') it is currently (or would be) drawing. My VDP emulator currently draws in increments of entire scanlines instead of slots, so the H counter would be awkward to implement. However, the V counter would be simple to emulate, albeit inaccurately (the real V counter continues counting upwards even after the last scanline has been drawn).

    Implementing this fixed issues in multiple games:

    OutRun's road not looking quite right.
    OutRun 2019 missing part of its HUD.

    I would have implemented this long ago had I known that I had not already done so. It was such a simple feature that I thought I implemented it back when I first wrote the VDP emulator, but I had not. Silly me.

    Vastly-Improved Mega CD Support
    For the first time, this project has received improvements from another developer! Devon has recently been learning the inner-workings of the Mega CD, and applied that knowledge to improving this emulator!

    Fixes and Accuracy Improvements
    It began with fleshing-out the stub LLE BIOS to handle level-2 interrupts and implementing a few stubs for the Mega CD's graphical capabilities. What sounds like relatively-unimportant busywork actually made a significant difference: with these changes alone, Sonic CD was finally able to progress past its title screen and into gameplay!

    From analysing the game with my 68000 disassembler, I knew that the freeze on the title screen was caused by the game waiting for an interrupt to occur, but I did not expect that interrupt to be related to graphical operations of all things! After all, the title screen itself was running just fine, despite its use of that fancy cloud effect.

    But this was just the beginning! Additional fixes rolled-in to properly-support the Mega CD's horizontal interrupt setting, allowing the game's water levels to run without crashing!

    Curiously, this feature suggests that the Mega CD's bus has a nasty hack in it to allow two specific bytes of the BIOS ROM's vector table to be overwritten. It may not be pretty, but it works.

    Also, a bug within the Mega CD hardware was emulated, allowing the game's FMVs to look correct:

    The way that this bug works is that DMA transfers which are performed from the Mega CD's Word-RAM suffer from a delay, causing all transferred data to be offset by one word. This is simple to emulate in the bus logic, just like the aforementioned horizontal interrupt override.

    New BIOS
    Along with these fixes came a whole new stub BIOS! Courtesy of Devon, this BIOS has been created by reverse-engineering the original Sega BIOS, accurately implementing a wealth of features! At the point, calling it a 'stub' BIOS might seem a bit harder to justify given how much more complex it is, but this new BIOS does not replace the high-level emulation that already exists. Like the original stub BIOS, this new BIOS merely handles tasks which would have been cumbersome to emulate in a high-level manner, such as invoking the Sub-68000 program and handling interrupts.

    Devon was particularly concerned with compatibility when creating this BIOS. Because of this, the BIOS also supports briefly displaying a splash-screen on boot:

    The original Mega CD BIOS features a well-known animation, showing the Sega logo. When this animation is complete, the BIOS does not unload its graphics. Because of this, software is able to use these graphics in various ways, such as in a transition effect or by using the font for displaying a menu. In order to allow these effects to be enjoyed, this splash-screen is needed.

    Sonic Megamix applies the Sonic engine's fade-out effect to the logo.
    CrazySonic humorously makes the logo mimic a reversing bus.
    The Bad Apple tech-demo stretches the logo.
    'Sega Loader' uses the font to display its menu.

    Emulation of the RF5C164 PCM Chip
    With Sonic CD being playable from beginning to end, the lack of emulation for various Mega CD components has become quite evident: the special stages lack their floor graphics, CD-DA music does not play, and PCM sounds are not audible. To begin addressing this, Devon and I have implemented PCM emulation!

    The RF5C164 is pretty rudimentary: it has 8 channels which sample PCM audio from the chip's dedicated RAM in a nearest-neighbour manner, and this sampled data is scaled by the channel's volume and panning settings before being mixed with the other channels' sample data. Compared to the YM2612, this is simple. There are still a number of edge-cases that need to be tested on real hardware, but otherwise this chip is now fully emulated.

    Emulate Level-3 Sub-68000 Interrupt
    Unfortunately, emulating the PCM chip alone was not enough to get Sonic CD's PCM audio to play. This is because the game's sound driver only runs in response to the level-3 interrupt of the Sub-68000 CPU. Without implementing this timer, you get no PCM audio at all. Without implementing this timer correctly, you get this:

    I expected this timer to be difficult to implement, due to how concurrency of emulated components works in my emulator, but it proved to be fairly easy. With this, Sonic CD's PCM audio finally worked... mostly.

    The reason for the speed being wrong was that the Mega CD has its own master clock, which differs from the one in the Mega Drive (50MHz instead of 52MHz). The emulator was not designed with multiple master clocks in mind, so Mega CD emulation was using the Mega Drive clock instead.

    After some refactoring and many headaches, support for multiple master clocks was added and PCM audio finally sounded correct!

    But there was still more to do: PCM is just one of two extra forms of audio that the Mega CD adds!

    Emulation of CD-DA Playback
    The PCM chip works by playing and mixing individual audio samples in real-time. This is old-fashioned, working somewhat like a MIDI file. On the other hand, CD-DA (Compact Disc - Digital Audio) is much more modern: it is a direct recording of music, just like an MP3 or WAV file. This frees the music from limitations such as the number of channels, juggling usage of channels with SFX, and ensuring that samples fit within a buffer. The downside to CD-DA music is that it is massive: Sonic CD's Palmtree Panic Present music is around 15MiB, which is 30 times larger than the entirety of Sonic 1!

    Sonic CD uses both PCM and CD-DA forms of music: using PCM for the music of the 'past' levels, and CD-DA for the music of the 'present' and 'future' stages. If I wanted Sonic CD to be fully emulated, I needed to implement both of them.

    CD-DA playback is a relatively high-level feature of the Mega CD: software simply sends commands to the sub-68000 BIOS to seek to and play audio tracks, with the actual process of fetching, processing, and playing the audio being hidden completely behind the BIOS. Since the BIOS is emulated in a high-level manner, this leaves me free to implement CD-DA playback however I want, without worrying about low-level details such as emulating a CD drive's controller chip.

    CD-DA audio data is raw 16-bit stereo PCM, just like a typical WAV file. It is not stored in any compressed or encrypted form, allowing it to be read by simply copying data directly from the disc image. The emulator's audio mixer just so happens to take 16-bit PCM as its native input format, making the process of "emulating" CD-DA as simple as a single file read operation.

    At the moment, features such as fading, fast-forwarding, and reverse playback are not emulated, but what is currently implemented is enough for most uses.

    The biggest problem with adding support for CD-DA playback was actually the file format used by Mega CD game rips: instead of everything being compacted into a single file, like a cartridge ROM, Mega CD games are split to at least two files - a BIN file and a CUE file. The BIN file contains a raw dump of all of the CD's sectors, while the CUE file contains a text-based expression of the CD's Table of Contents. The CUE file format is strangely lacking in a concrete specification, requires slow and complicated text processing logic, and also outright requires the use of file IO. That last point especially is a huge problem, as the Emscripten port of the emulator is unable to use file IO due to running in a sandbox. This would not be a problem if both files were simply combined into one.

    Another issue with CUE files is that it is possible for the audio tracks to be extracted from the BIN file and be stored as their own standalone files. These files can be encoded in the WAV format, Ogg Vorbis format, MP3 format, etc. This creates a huge burden for the emulator, as it needs to support every feasible audio format in order to support as many types of Mega CD disc rip as possible. This requires pulling in numerous audio libraries like libvorbis and minimp3, each one bringing its own licensing terms, executable bloat, and portability concerns.

    In recent years, a new format has emerged: MAME's 'CHD' format. This format combines everything into a single file, while applying lossless compression to reduce its gargantuan size. Unfortunately, the other problems that I listed above apply to this format at well: being such an elaborate format, and leveraging multiple forms of compression, numerous libraries are required in order to decode CHD files. Given that clownmdemu's core is written in portable ANSI C while libchdr and zstd are not written portably at all, I really, really do not want to use them.

    For now, the emulator supports the 'BIN+CUE' format. This still requires file IO which will prevent it from working in the Emscripten port, but it at least does not require additional libraries.

    To help with supporting the wide variety of CD formats, I have split the emulator's CD code off to its own library - clowncd. This can be used for more than just emulation, as shown by its test program, which can extract ISO and WAV files from a BIN+CUE rip.

    6-Button Controller Support
    Up until this point, the only type of controller that was emulated was the original 3-button controller. The emulator has now been extended to support the later 6-button controller, allowing software which supports it to be used with a more natural control scheme. Some games have cheat codes that are only possible to input with a 6-button controller.

    This controller was considerably more complicated to support than the 3-button controller, due to it internally featuring both a counter and a timer that managed which buttons could be read by the Mega Drive at once. This is because the 9-pin port that the controller is plugged into is not enough to expose all of the button data at a single time. This was already a problem for the 3-button controller, which was solved by splitting the buttons into two groups and using one of the pins to select which group to be read with the other pins. The 6-button controller extended this, exposing the additional buttons after the 'select' pin has been strobed a certain number of times, and resetting this number after some time has passed. Emulating this required making the controller emulation logic able to measure time, similar to the CPU emulation logic.

    Add Per-Operator Frequencies and CSM Mode to the YM2612 Emulation
    The YM2612 has a strange feature, where the third channel can use four different frequencies instead of one. This feature is heavily used by Streets of Rage. A related feature is the so-called "composite sine mode" (CSM), which triggers a 'key-on' and 'key-off' of the third channel when Timer A expires. According to Chilly Willy, this feature is intended for speech synthesis, but no official Mega Drive games appear to use it.

    When I created the YM2612 emulator, I did so in a very object oriented manner: the FM chip as a whole was an object, which contained 6 channel objects, which contained 4 operator objects, which contained a phase generator object and an envelope generator object. This was somewhat overkill, as such heavy encapsulation meant that objects did not share state, even if the components of a YM2612 actually did. Ironically, this actually worked in my favour, as it meant that each operator already had its own frequency, albeit one that was always set to the same value as the channel's other operators. So, to add support for the third channel's per-operator frequency mode, all I had to do was allow each operator to be assigned a unique frequency.

    In the future, I hope to refactor the YM2612 emulator to be less encapsulated, allowing for the elimination of redundant duplicate state.

    CSM was comparatively simple to emulate, however it is not entirely accurate. All that is needed is to call the key-on function twice when Timer A expires - one to turn it on, and the other to turn it off. A real YM2612 merely queues these operations, and performs them during the envelope generator update process. This delay could have side-effects which are not currently recreated in my emulator.

    Implementing both of these features allows Nemesis's CSM test ROM and Streets of Rage to work correctly.

    Add SSG-EG Emulation
    SSG-EG is yet another curious feature of the YM2612. I was able to implement it rather painlessly thanks to Nemesis's wonderful documentation. The way that it works is that it builds upon the YM2612's ADSR envelope generator, allowing the envelope to be played repeatedly, optionally with mirroring, inversion of the envelope's attenuation, and with the option of having the envelope hold its final value when it ends. It is a bit of an esoteric feature, but many official games do use it.

    A couple of games (an Olympics game and a Beavis and Butthead game) use SSG-EG in a way that is officially undefined, causing the YM2612 to behave in a buggy manner. In order for these games to produce audio correctly, SSG-EG must be implemented in a way that is as close to a real YM2612 as possible. Fortunately for me, this is exactly what Nemesis's documentation helps to achieve.

    SSG-EG is actually a leftover feature from the YM2612's predecessor, the YM2608. In that chip, there was a "software-controlled sound generator" (SSG) module alongside the FM module. The YM2612 only contains the FM module, but since SSG-EG was part of the FM module and not the SSG module, it is present nonetheless. The SSG module and SSG-EG were necessary for providing compatibility with the AY-3-8910.

    Fix Sonic 3's Competition Menu Music and Contra: Hard Corps' Snare
    Ever since YM2612 emulation was first added to this emulator, one of the instruments in Sonic 3's Competition menu music was really loud. This was the only FM audio issue that I knew of which was caused by an actual bug rather than just a missing feature. That changed when BlastBrothers put together this huge list of audio issues, ranging from missing features to new bugs. In particular, Contra: Hard Corps has an FM snare drum that was playing completely differently to how it does on a real Mega Drive. I figured that, by fixing the underlying bug which was causing one of the inaccuracies, there was a chance that it would also correct the other inaccuracy.

    To debug this, I began modifying the Sonic 3 disassembly to produce a ROM with a slightly different instrument, altering various settings within the instrument until the ROM sounded identically in both my emulator and a real Mega Drive. Eventually I discovered multiple ways to make the two sound the same:
    • Change the sustain rate to 0 to match the decay rate.
    • Change the decay rate from 0 to match the sustain rate.
    • Change the sustain level from 0 to any other value.
    The decay rate and sustain rate are two variables which control the ADSR envelope. The sustain level controls when the envelope switches from using the decay rate to using the sustain rate. These three modifications suggested that there was something wrong with how the sustain level was being used: since the instrument has a sustain level of 0, the sustain rate should be switched to immediately, leaving the decay rate unused, and yet the instrument would sound wrong in my emulator unless the decay rate matched the sustain rate. This meant that the emulator was using the decay rate when it should not be.

    After some trial-and-error, I found the cause of the bug: the documentation which I had used to develop the YM2612 emulator was wrong!

    As described in a previous blog post, I relied heavily on Nemesis's documentation when creating the FM emulation. This documentation claims that the envelope generator is not updated on every sample; instead, it is updated every several cycles, depending on what the current rate is. Here is the pseudocode that was provided to illustrate this:

    For each operator
       //Read the shift value from Table 1
       counterShiftValue = counterShiftTable[operator.rate]
       If (globalCycleCounter % (1 << counterShiftValue)) == 0
           //Check the current cycle we're up to (0-7)
           updateCycle = (globalCycleCounter >> counterShiftValue) & 0x07
           //Read the next attenuation increment value from Table 2
           attenuationIncrement = attenuationIncrementTable[operator.rate][updateCycle]
           //Update the attenuation
           //The attenuation update process happens here. More on that in the next section.
    Next operator

    Because my emulator does just this, there is a large period of time before the emulator checks the sustain level and switches from the decay rate to the sustain rate. In reality, the YM2612 updates the envelope generator on every sample, and the current rate simply affects how much the envelope is progressed. This can be proven by examining Nuked OPN2, which is a transcription of the YM2612's circuitry to C.

    With this problem solved, both games sound correct. Just as I had hoped, fixing one inaccuracy also fixed the other!

    Perfect Frame Advance Rewinding
    Because the rewinding feature predates the frame-advancing feature, the former was not designed with the latter in mind. As a result, when used together, they exhibited latency: when toggling between advancing forward and advancing backward, it would take multiple frames for the change to actually take effect. There was also an additional problem where, when advancing backwards for the first time, it would not advance at all, merely duplicating the previous frame. The cause of this had to do with how the rewind feature's ring buffer worked; particularly how it was read from and when it was updated. By adjusting these details, the latency is eliminated, making frame advance rewinding work perfectly!

    So, here it is - one giant update. As always, there is plenty of work left to be done, but hopefully what is here already is enough to be useful.
    Last edited: Apr 1, 2024
    Pacca, ProjectFM, JGamer2151 and 2 others like this.
  19. peachy

    peachy Newcomer Trialist

    May 31, 2023
    There's another format for Sega CD game files (tho idk how common this is) where the game data is stored as an ISO file w/ the CD soundtrack being stored externally. Loading the ISO itself in this example didn't work, it might be fixed by checking for MP3s in the same folder? scd1.png
  20. Devon

    Devon I'm a loser, baby, so why don't you kill me? Member

    Aug 26, 2013
    your mom
    ProjectFM and peachy like this.