I was mind blown away about how two mixed opposite audio samples do cancle out eachother because i was listen it in mono ion my iphone,trough i could still hear some smaking sounds ,however once i did put on an earphone i could clearly hear the music ,alltrough somehow they did let me able to even hear that music on 4,20 even in mono, i don’t know how they did this trick,now i tested it by recording the mixed audio into izotope RX7 and strangely enough i do could somehow hear the music very very soft,but with the music balance tool i was able to saperate the drums from the other instruments and put them on saperate channels,boosted up the volume and bam,now i could clearly hear those saperated sounds,i was mind blown away ,so even 2 opposite mixed audio wich cancled eachother out can be recovered with sound saperation algorithms found in the music rebalance module of izotope RX7,just wow,incredible it’s just amezing such impossible thing👍
Funny thing is that you can hear a little bit of the "non-effect" version in the DS version. I say a little bit as it seems to play alongside some of the effect version.
6:30 wow, I recognized that grating, high-pitched tone just as quickly as the wind itself! were early SNES emulators generally unable to do the pitch mod effect? That's the only reason I can think of that I would find that sound so familiar.
Quite possibly! Obviously, the earliest emulators were trying to figure out the SNES in general, and sound was one of the last features for hackers to figure out... but I don't have any idea how common it was for early emulators to ignore pitch mod. That's 20 years of software history to dig through right there
I can confirm that the early zsnes versions had that issue, for example, with the wind howl in FF6. How early? I can't pinpoint a date, but I can tell you I was downloading roms and emulators with a 56k modem, so I'd guess late 90s / early 2000s.
@@QuintusCunctator Yeah, that specific sound effect of wind in Squaresoft games was a huge turning point in early SNES emulation, before we started getting into "can it do X specific chip", because it was real easy for anyone who had played the game on a console to recognize the sheer "that's not right" factor. It would have been in the first half of the 00s when it clicked but I can't remember the specific year.
I also believe that some issue may have cropped up when porting the music to the DS for Chrono Trigger's 2008 remake. The atonal wind sound is definitely present, but moreso in conjunction with the high-pitched tone. It's a strange discrepancy especially considering how tight of a remake that game is - definitely not done by the same teams who end up making all those crappy phone versions that are then put up on Steam for $20+, that's for sure!
That pitch modulation example reminds reminds me of when I played FFVI on SNES9x around 1999. The wind sounded exactly like your demonstration of disabled pitch modulation.
Hey, I absolutely LOVE that little countdown timer you use to tell people when a piece of onscreen text will be going away. I genuinely wish more RU-vidrs would include something like it.
Glad to see someone gets some use from that! It's a feature I wanted to see on other videos when large walls of text appeared, so I started including it in my own.
Well it is indeed well explained in this video albeit oversimplified🙁, apart from those freaking stupid annoying enthusiast sarcastic generic jokes in it,damnit😡🛑
Ahhhhhhh... I WAS WAITING FOR THIS~!! The inverse-audio thing is basically what allows surround-sound support in certain SNES games, as Dolby Surround back then took both channels and any waveforms that were literally inverse in one channel that is right-side up on another were instead put into the back speaker, while everything else was on the forward speakers.
@@T3KNUG3T5 Well, it doesn't sound like Q-Sound has an actual physical back speaker, but rather just 2 speakers that try to create a certain binaural image.
Yep, Dolby pro Logic. The first dolby form. Left (wave on left only), right (wave on right only), center (wave same on both channels), and rear (wave inversed on both channes. Rear channel was also very frequency limited, no highs and no lows, just like BOSE, lol
I'm working on SNES programming right now, and the sound stuff was very hard to figure out (I knew the basics for a while, but the specifics on programming are a mess). I wasn't expecting this video to go much actually in depth, but you exceeded my expectations pretty quickly. Great work!
4 года назад
The SNES Port of Prince Of Persia uses pitch modulation across the entire soundtrack, as well as vibrato.
6:29 Funny enough, when I played Chrono Trigger for the first time, it was the port on Final Fantasy Chronicles for the PlayStation. Any scene that had the wind sound effect playing pretty much sounded exactly like that.
Also, side note about the 8 channel limitation: Although the snes could indeed have 8 stereo channels of sound, many games reserved anywhere from 1-3 channels for sfx. Typically these are channels 6-8, which is the format Super metroid uses, as playing a sample in channel 6 will play over a SFX that would be assigned to that channel
This is a common technique for plenty of consoles / old games, including stuff on the Genesis and Amiga In fact, I recall that Jeroen Tel wrote a C64 tune using 2 channels, with the 3rd one reserved for sound effects. It goes way back!
@@GSTChannelVEVO - Actually Jeroen wrote all of the BGM for RoboCop 3 on the C64 (wich of course doesn't count the title music) using only one channel, and it sounds waaaay more complex than it should.
64KB isn't so bad - with smart use engineers squeezed out some amazing stuff. The most important part of most instruments is the attack. The "chuff" of a flute, the "pluck" of a harp, etc. These attacks are milliseconds long. The rest can be a tiny single cycle loop that can be pitch or volume modulated. Really clever engineers would create single cycle waveforms that contained chords, such as the organ in Plok. That maximized sample memory AND helped the music sound richer since it only used a single channel, leaving more channels for more instruments. (Crazily enough Plok used only 5 channels for its music!!) You could synthesize a kick drum with a single-cycle sine wave and a good custom audio driver. The most efficient way to do sound effects was to modulate instruments, controlled by the main CPU, which saved RAM on the APU by storing the sound effect data in main RAM. Earthbound did all of its sound effects this way, leaving more space for all those samples from pop music. (Also I think reverb wasn't used.) The hardest stuff to optimize are snares and hihats, which have a lot of high frequency content, so when you're listening to ripped SNES samples you'll often hear a barrage of blips and glitches and a couple of very clear drum sounds sticking out. On the reverse side, heavily downsampled snares sound good when pitched up, adding clarity to the mix - this was done in the Kirby games, Earthbound, and a Japan-only game Wagyan Paradise. Why do I know all this stuff? Listening to a ton of SPC music and being hobbyist game dev / musician... I LOVE optimization. I find it endlessly fascinating... the closest thing to magic.
or platform games already enough 64kb, but for fighting games it's too bad the voices are muffled because of the low quality and cut to fit the spc700 aram, street fighter 2 is an example
The part about the inverse sound blew my mind. My phone is mono so I heard nothing, then switched to my tablet and bam, literally sounded like it was inside my head.
3:41 actually you can use this trick to create surround sound if you have the correct equipment to play it this example you give, in a surround sound equipment plays only in the rear channel some SNES games have sound made in surround, and some of them are very well made, like Secret of Mana 2
A correction about the pitch modulation. The OPL2 FM modulation and SNES pitch modulation don't work the same. The YM3812 uses phase modulation, not conventional FM modulation. This is why those FM synthesiser apps don't sound like a YM3812 or YM2151.
The Genesis is synth based. All you need is the patch data + song data, and a driver that converts that data into register writes for the YM2612 (and SN76). There is the option for playing samples on channel 6 of the YM2612, but that's usually more of a problem with cart space than memory, as the sample data is thrown at the DAC a few bytes at a time (as opposed to being stored in a special RAM area and being read when needed a la the SNES) I don't know / remember enough about the Sega CD or 32X to know if memory limitations were an issue there, but I would guess "yes" (at least for the former)
Chris James The main thing IMO that wasn’t explained (probably because it’s kinda complicated) is the gaussian interpolation of the samples. This is IMO the main cause of the muffled sound on the SNES with low quality samples making it worse. There were some techniques to alleviate it, but I don’t think a lot of games used them. The Genesis only used samples for a single channel so it was mostly percussion and a few effects. You could also (with a few caveats) stream audio into the audio chips which was used for the famous MOD playback in Toy Story where the 68000 CPU is mixing 4 channels in software. FM instruments are really small because there are only a few parameters that change like algorithm used, operator volumes and pitches.
@@GSTChannelVEVO Ah right, even if you had upwards of 12 FM patches to cover all the instruments you need in a song, it wouldn't matter as far as memory, forgot how tiny those were.
The guys who are putting some of my recent tracks into Rock 'n Roll Racing said it is an issue for that sound driver at least (which kinda sucks though); I don't know the details but there is a strict limit on file sizes and instructions between when the commentator voice samples are playing.
@@GSTChannelVEVO The sound chip on the Sega CD funnily enough had similar specs to the SNES (64k RAM and 8 bit 32khz playback), but I'm not sure how much impact that had and there's not a lot of music that used it outside of Sonic CD past
5:45 the SCSP (Yamaha YMF-292) in the Saturn (and its DreamCast variant, which may be different) has 32 operators that (with limitations) can modulate each other. Each operator is either put in sine wave or PCM mode.
I see at 2:30 you are showing the C700 VST, it's a great tool for DAWs and gives a lot of info, but it's usually better for display purposes since most spc music can be written or decided to .It format
When working on mixes, if I need to write a new section to transition between two songs, I can load the songs up in C700 and quickly sketch out a transition. It's a nice shortcut compared to SNESMOD since I get the authentic sound without having to convert + render from SPC, though this may be a very specific use case here. ;)
Oh god the moment I noticed it on my phone I ran full speed to my PC so I could listen with headphones because oh boy I didn't expect to see this today
Excellent video! I love these so much! Its really funny how fm sampled waveforms on snes can work and sound much better than most regular sample forms lol. Yet there only 2 snes games that used them.
Will have to rewatch this to understand everything better, but following that leap from the Sega Genesis to the Super NES, I hope you'll make one sometime to the Playstation 1. I came across the Rockman Complete Works soundtracks before (Shadow Man's theme was my favorite), which reminded me of how far superior to my SNES and Nintendo 64 the audio in some PSX games were that a friend of mine used to own, and couldn't help but think, "This would never fit into a ROMhack of Earthbound as a possible battle theme." (That's when I got the idea to aim instead for a whole new game from scratch.)
3:53 DUDE THATS CRAZY! I listened to that music, then I went to my phone’s settings and switched to mono, listened again, and the sound was gone! That’s trippy as hell.
My limited understanding of Saturn sound programming is that it's 32 channels of sample playback or fm control oscillators, each of which could then be used in an arbitrarily complex FM operator (basically a roll-your-own-fm-op setup, not clear on the details), plus effects on top. I'm not sure the limitations of it, but it looks to me like a ridiculously powerful piece of synthesis hardware. Granted, most people used it in a relatively straightforward fashion, because it was easier to get consistant results from sample playback, and dividing the channels to make FM operators made sound synthesis and sound effects harder for most game dev purposes. If I weren't terribad at coding and Saturn homebrew wasn't crazy difficult, I'd love to put together a hardware/software solution to convert the Saturn into a ponderously flexible synthesizer module
Hi GST Channel! I really love the information in this video, so I'll gladly subscribe! I'll be taking a look into SNES SFX Hacking for an EarthBound ROM hack :)
I believe that the SNES could theoretically store a short loop of slightly muffled CD audio that would basically just sound like an MP3, but there would be no other room for other data in the game. Correct me if I'm wrong because I might not have any idea of what I'm talking about.
kind of! I'm not sure how much music you could fit on an average cartridge, but as a technique, you could definitely stream muffled CD audio to the soundchip. there are even a few instances of something like that, like the six second piano loop in Jurassic Park. check out my video titled "Crazy Sample Usage in SNES Music" where I go over that.
@@GSTChannelVEVOI actually checked that out, and it states that it uses half of the ram. So, does that mean that the max would be 12 seconds? Of course, this is without chopping up and rearranging the samples. Again, SNES cartridge sizes vary so this limit probably does too.
@@breadtheloaf-yf5tc the max you can fit in RAM is about 12 seconds, unless you muffle the sound even more. though it's also possible to swap out the contents of the RAM as your audio is playing, enabling you to go far beyond 12 seconds (and/or with less muffled audio), but then cartridge space becomes an issue
Try to make a video explaining the modulation in SPC700, although it was very basic there are many people interested in knowing more about that. It is one of my most watched videos of my channel. On sound modulation, make some arrangements with SNESMOD dirver or custom sound by Mukanda, but it is far from exploiting the sound capabilities, as did the RARE or Square games sound engines.
every time someone captures youtube's layout in a video, it becomes a time capsule, usually of better times. I'm pretty blase about rounded corners but what misguided soul truly wanted rounded numbers, I wonder 😔
That wind sound without the pitch shifting. Old emulators didnt do that correctly, back then? Explains why the sound effect of Megaman X, when he collects a new power up and throws his fist in the air sounded so wrong on emulators back in 1998.
Short answer: there are some default instruments but not everyone had access to them. Longer answer: I explored it in this video: ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-lA3LzyH6cEg.html
In the end, making music for SNES is way easier than making Genesis music, imo. Even with 48KB RAM limitations. Or I'm just used to working with tracked musics (MOD, S3M, XM, IT) than FM synth sequencer, and granted I used SNESMOD to convert IT module to SNES-compatible SPC data.
It is important to note that with the Sega Genesis/Mega Drive is that many Japanese computer systems used Yamaha sound chips in their systems, as well as many arcade machines also using Yamaha sound chips. Many of the specs on the Genesis were picked due to Sega wanting to make sure their arcade games could be programmed easily for the Genesis. By 1989 (ie. the release of the Genesis/Mega drive), a lot of Japanese computer developers also used a computer only manufactured in Japan called the Sharp X68000 which was easy to program for and had top of the line capabilities (The X68000 cost over $8000 (US currency) at retail if accounting for inflation), and then the Genesis specs were cut down to be affordable for the general public. The familiarity of Yamaha sound chips meant that even though it was a different serial number, the programming wouldnt be too hard to figure out from one Yamaha sound chip to the other.
hey! i don't mind the muffled sound. but not all of the snes games voice clips sound muffled. if you have played super tecmo bowl on snes, they sound like they're in a hallway close to somebody.
Retro Gaming EXPLAINED has made a video about the SPC700. Also, I have a question just for curiosity: Why did you change your channel from Mega Drive music to Super Nintendo music?
I want to explore ALL of the game music. I started with the Sega Genesis because that's what gave me the idea to explore game music like this. Eventually, I was satisfied with my exploration on the Genesis and wanted to focus on another console. I left it to a poll, and the SNES won. In the future, I'll shift my focus again to some other console.
I'm having an issue with my Snes, when I play Street Fighter Alpha 2. There are music but I can't or don't hear the announcer or Ryu's hadoken sound effect. How much does it cost to fix it?
Woow you have things well explained,,the lowpass filter,for filter the and the reverb filter are interesting features,but i wish you explained more about the guessian filter,now am also curious how the snes turns those BRR/adpcm samples somewhat back into 16bit ,if game devs rather used lower quality samples,i wonder if they also could use 1bit & 2bit adpcm samples to save up more space,would be interesting,now likely with the flexible sound driver engine it was possible to get around the 64K limit by streaming sections of audio samples one after another to the snes,allowing for true & full 16bit audio quality👍
Look up the Nyquist Theorem. Basically, you can get away with fewer sample points (lower sample rate) with *no* loss to the waveform, other than limiting the high frequencies. If you have a waveform sampled at a lower rate and you want to play it at a higher sampling rate, you need to interpolate new points. The gaussian filter does that. A proper reconstruction would use a sinc filter, but that requires more hardware and the only benefit is a little bit less reduction in high frequencies. Either one takes several sample points and then generates the proper sample for any fractional position between the two middle samples. The BRR was a nifty scheme. Basically each block of samples could be run through all the different filters during compression when developing the game, and the one that yielded the least distortion could be chosen.
@@gblargg i make my only examples at 16khz and they are flustered is it possible to solve this? i wanted to learn how brr compression works to improve my samples is there a document explaining well how this compression works?
Just the documentation on how the SPC-700 works. I don't have experience encoding samples or making music. There might be some tools to help with making SPC music.
I don't think so. I wouldn't remember but I don't see left/right channel-specific flags for the echo buffer or individual channels. Of course some people set one of the voice's volumes negative so you get an out of phase effect (which sounds pretty bad, especially on headphones). Surround killer to the rescue.
Retro Game Mechanics Explained has made the SPC700 video now, if you're looking for it(LOL): ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-n5eTOGZdnTU.html
Sort of! an LFO is (as it says on the tin) just a low-frequency oscillator. it's common to route a sinewave LFO to the pitch of a synth to get vibrato. if you played a really low sine wave in the pitchmod channel on your SNES, it'd have exactly the same effect: vibrato. however, you can drastically increase the frequency of that sinewave, which causes the "vibrato" to become "new harmonics" --at which point it's just called frequency modulation synthesis.
sort of! I think this part of the video wasn't as clear as it should have been. I took the source code of an existing emulator and made some changes to it. *that* emulator was written in x86.
Worded differently, I try to make sure that the audio in every SNES-focused GSTMIX video is able to be played back on the 8 channel hardware in the SNES.
@@GSTChannelVEVO Sorry, I tried to make a pun on what you were saying about your RU-vid channel in the beginning and the apu channels, but it came out worse than I expected
probably this: wiki.superfamicom.org/ I've also seen SNES homebrew people thank RetroGameMechanicsExplained as a reference, lol. as for your second question: iirc, NES sound was not supported, though I do remember there was a noise generator somewhere in the SPC700. but also, it sounded really different from the NES.
5 лет назад
Ah, the SPC700. A great sound chip, if properly coded. The 64 KB memory limitation explains why SF2 voice samples were restricted to less than one second. Looking forward for the next videos.
Hello, I have a 1chip outputting very low audio from the dac converter that goes low all the way to the rf and composite, caps measure all good, voltages are ok, my local tech says he's never seen such problem, do you have any idea what could be causing this(tested on different tvs various cables and jumpered everything tried all possible fixes with jumpers)
it's a basic sampler. you could use it as a wavetable synth by storing several waveforms and playing them in succession, but it lacks the ability to natively switch between each waveform, which is kind of important for a wavetable synth.
as in, telling the C700 "play this sample" and then streaming an entire song just ahead of where the C700 is playing? I believe this is possible but I'm not sure I've seen it done. the Tales of Phantasia opening comes to mind, but I think that chops it up a bit more carefully than the "throw constant data at the DAC" megadrive method.
Wow, the SNES sound chip is SEVERELY DISAPPOINTING! The Amiga's Paula audio chip is BETTER, and OLDER, too! I'd like to see the SNES attempt 14-bit PCM IN STEREO, to be honest!
Paula is its own beast! I think the SPC has a few mild advantages (delay and soft panning supported in hardware... and I don't think paula could do FM natively) but it's not exactly a fair fight going against the Amiga :P
4:15 The SNES uses a 65816 as its main CPU, not x86. Other than that, you actually have more in-depth knowledge about the SNES audio hardware than Nerdwriter1 does!
this comment scared me for a moment, lol you're 100% correct, but I was talking about editing an emulator at that timestamp. specifically, I was editing the source to Alpha-II's in_snes, which was made in x86 assembly.
you're right! I think I missed that in the video. basically, there's a register that will cause a channel to replace the waveform of the currently playing sample with generated random noise instead
The SPC was quite something. Some games (mainly Capcom and Square) made great use of it. I legit wonder if it was possible to use it to replicate sounds from FM chips, like the Genesis'
It's super easy to sample FM sounds. For fun, kulor and tssf have recreated some Sonic tunes in SPC format, if you want to google those. There's also the soundtrack to Super Tekkyuu Fight but I don't think anyone has replicated FM sounds using the Pitch Mod feature. Usually that's used for more complicated effects.
I was looking up what adpcm was which the S-SMP uses. It's 'adaptive differential pulse-code modulation' which means it predicts the gaps in low frequency samples and was developed by Bell Labs for telephony. The samples are initially 4bit and upscaled to 15bit. Although the S-SMP can output 16bit 32Khz but that would have to over only 1 channel. The frequency output is lowered by half for four channels down to 4Khz for all 8 channels running simultaneously correct me if I'm wrong. It will be tough to find interesting tunes on the snes. The low fidelity murders most music. The samples are such low quality. It is idiosyncratic, but the problem with too much noise, gaussian interpolation, reverb and echo make the catalogue less than stellar and pretty samey. Get used to a whole lot orchestra hits and breathy strings. Bah. Another handicap other than low memory the S-SMP (sound unit) had was it did not support mod file format. So you cannot compose a chip tune on the snes without first writing your own tracker program which is where alot of 16bit games get their distinctive music through the methodologies of tracker composition and use of the 'echo' effect to generate the fluttering sound. So alot of the music on the snes is composed on midi and as such suffers from a lack of expression and a plodding type of sound.
even with the limitations I think it's a great sound chip look at it imitating the ps1: www.smwcentral.net/?p=section&a=details&id=19014 N64:ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-0vDnn2Yu8zU.html
you appear to be describing the N163, an ASIC that appeared on a handful of (Japanese) NES titles. The S-SMP is consistently 16-bit 32kHz. as for MOD file format support: this is a driver issue, not a hardware issue (mostly!). Some drivers did exist that emulated the MOD format, but with extra bits that took advantage of the SNES's features. The lack of proper dev kits was a common complaint among musicians, I've noticed. as for the samey music: there's plenty of stuff that expands into other fun genres! I suggest checking out GSTMIX11 ;) (granted, that still has orch hits in the form of rave stabs, but i like those quite a lot)
@@GSTChannelVEVO I understood most of that. Not sure where I got my frequency info or if it's right or not. But mod format was not natively supported, whereas the Mega Drive 1 pcm channel did support mod and regular pcm, making trackers easily useable for the mega drive as opposed to the snes requiring drivers to expand the sound chip's compatibility. The familiarity with tracker software of the musicians is the main reason FM music is overall, more expressive than pcm format midi composed music which software such as Cubase lacks the ability to easily program arpeggios and portamento, utilising more channels of sound and with a higher degree of complexity than the S-SMP provided with the adpcm samples at 4bit. Conversely, all Mega Drive fm music is 16bit 56Khz and only as good as the composer and driver used, whereas the S-SMP is tied to a particular range of sound by the nature of the limited memory, unlike the Amiga which could generate crisp and clear music over 3 channels with one left for sound effects. I like some SNES music but I do not like the Castlevania 4 sound track.
the genesis didn't support mod playback natively either! Though I think you're referring to the mod player that appeared in the Genesis version of Toy Story? That was a very clever programming trick, not native support. Of course, some people did write trackers for the Genesis, as heard in Jesper Kyd's music, but that's a rarity. a sort of MIDI method was most common on the Genesis (see my video, "how to make sega genesis music in 1994") Note that every piece of hardware requires some sort of driver. that's what the name refers to: code that drives the hardware. more advanced drivers can push more quality out of the hardware, but don't confuse the limitations of the driver with limitations of the hardware. similarly, don't confuse the limitations of the driver with the limitations of the composer :P Incidentally, trackers for the Amiga aren't exactly "native" either. they're just the most intuitive and straight-forward way to work with the hardware, which expanded into other hardware architectures quite naturally.
@@GSTChannelVEVO Well I read somewhere the genesis supported mod playback on the pcm chip. But what I meant was tracker musicians could use existing tracker applications to compose music for the ym2612. European musicians were used to home computers like the C64, Atari ST and the Japanese were using X68000's and things using trackers to write chiptune music. It's a weird way of working but it is the only real way to get that specific sound, created to get the most out of simple fm chips but easily applied to pcm chips. Using GEMS to write music is basically input from a keyboard and therefore midi. Midi music is fine but you cannot apply envelopes to the instruments with pcm music with as much impact as you can with an fm instrument which can have all of it's features altered over time such as cutoff and resonance to apply impressive sweeping effects. The problem with the S-SMP is 64Kb is woeful. The 4bit adpcm expanded samples don't have enough depth in their limited range to achieve what the chip is claiming to be capable of. The result is something quite tinny and but also strangely hollow and unclear. I like some music composed for the SNES but it's limited clarity makes it something of a oddity amongst sound chips. It's sound is synonymous with the snes but for the majority of games, they suffered from the limited instruments on the sound track and lower quality sound effects. There's only so much you can do with echo and reverb but I don't know what people used to compose snes music with. I would assume an Atari ST with Logic or something.
