How Do You "Make" An Instrument?

[reversalmushroom]

I am refering to computer-generated instruments. What programs are used to create them? And for that matter how were console-generated instruments created? I'm aware that game developers pulled these sounds from the console's hardware, just like with the NES, but I'm curious as to how. How do you program an instrument? A lot of my favorite songs are from the SNES, and partly because of the sounds of the very instruments themselves; I'm amazed at the quality diversity of sound that developers got out of SNES hardware (though that's probably because I encoded my WAVs using Super Jukebox 3.3); Gradius III, Megaman X, Super Bomberman 2, Super Bomberman 5 and Super Mario RPG are but a few examples. I don't know about you but I think that synthetic instruments of today couldn't touch the SNES, and yet computers today have far more potential than they did back in 1991. My reasoning is that if the SNES could create something so sublime, that state-of-the-art technology could create something unimaginable.

SNES game-rips sound mostly the same, but creating them through Super Jukebox 3.3 gives them a clearer, richer, sharper, and all-around better sound. And it has to be 3.3. If you use 3.2 you might as well be using SNESamp, and anything below 3.3 must NEVER be used, or else 3.3 will forever sound like 3.2 and all other means of SPC conversion (I don't get it either). And it doesn't and can't work on certain computers. Now I know what you're thinking (Super Jukebox, ewww!), because it's extremely unstable and crashes if you so much as look at it wrong, but Super Jukebox THREE POINT THREE gives SPC files a sound that you simply don't get with other programs. Dare to compare:

http://www.4shared.com/audio/gpfzF7NT/G ... oss_2.html
http://www.4shared.com/audio/1_QG7WPe/G ... _Base.html
http://www.4shared.com/audio/Z8RPlbSi/M ... anger.html
http://www.4shared.com/audio/B5sz6-xK/M ... eleon.html
http://www.4shared.com/audio/p9Ud6Hsu/S ... M_5_2.html
http://www.4shared.com/audio/HaaWlQTy/S ... rdly_.html
http://www.4shared.com/audio/ST2Z2NEZ/S ... _Stat.html
http://www.4shared.com/audio/3I_3dMVw/S ... _Bomb.html
http://www.4shared.com/audio/0eQeRWIa/S ... arade.html
http://www.4shared.com/audio/lYLnOd1V/S ... Dream.html

PS - Does anyone else really like SNES music or even the sound of it?

[DC64]

Im okay with the sound,

I don't know about you but I think that synthetic instruments of today couldn't touch the SNES, and yet computers today have far more potential than they did back in 1991. My reasoning is that if the SNES could create something so sublime, that state-of-the-art technology could create something unimaginable.

You have a point, there is "live soundfonts" but I am not 100% satisfied it sounds like it should.
(nobody has gotten a French Horn right)

(Back to the SNES)
Reading here can give some info. Wonder if this helps.
http://en.wikipedia.org/wiki/Super_Nint ... stem#Audio
http://en.wikipedia.org/wiki/RF_modulator
http://nesdev.parodius.com/
http://en.wikibooks.org/wiki/Super_NES_ ... 0_programs

[Bregalad]

Well to make an instrument there is basically 3 ways of making them :
1° PSG : Use simple waveforms such as square waves, triangle waves, saw waves, etc.... The NES does this, so does most of '80s computers and consoles.
2° FM synthetis : I'm not sure how it works but basically it takes a sine wave, distort it by frequency-modulating if in the audible range and it creates another continous waveform. Has the advantage to be simple to do in hardware and takes almost no memory. The molulation parameters should vary in real time for interesting effects. This was used in the late 80s and early 90s. The megadrive does this.
3° Sampling : Use recordings of all instruments you want to reproduce. This is the most common method since the '90s. The amiga was probably one of the first popular computer to be able to do this. It's also easy to see you can record samples from the first two methods to simulate them, but the promlem comes to how many memory it takes to store all those instruments.

Also note that digital filter effects can be applied to all 3 methods for additional sounds to simulate more complex effects.

The SNES, simply put, uses method 3 like does all modern platforms, except that it's limited to about 6 seconds of 32kHz samples at a time in memory (and the replaying program and echo buffer also takes up space so in fact it's less than that, probably only 4 seconds). There is really nothing special about it - if the SNES can do it, so can all modern platforms.
On the other way, if you want to convert any sequenced music to be played on the SNES, you have to make sure it only uses 8 voices and that all samples fit in memory (that is only 4-6 seconds at 32kHz).
The only thing specific to the SNES I think is the way it can use echo, while modern platforms uses typically reverb which is a more complex type of echo. Echoing in the snes just repeats the sample quieter every so milliseconds, so it's really simple to implement, and sounds differently than reverb which simulate a true room. So that might be why you love the "sound of the SNES" so much ?

[reversalmushroom]

How were those samples (the one's used to become the instruments of SNES games) created? They don't sound like someone recorded a real instrument being played. If they did, then the SNES's music would sound like the next-gen console instruments that are recordings of real instruments.

[Mark7]

I believe most instruments on the snes originate from high quality soundbanks which were used in keyboard etc. Then you should ask yourself how those intruments were created. Well, some might be synthesized but others might come from a real instrument recordings.

But why don't they sound real on the snes then? I blame the sound memory of the snes. It is only like 65kb in total. So only extremely small samples can be be used, OR extremely low quality samples (very low sample frequency). So that pretty much kills all the "realness" of the instruments.

If you take amiga for example, like Bregalad was talking about, you will notice those instuments sound a bit more real than the snes (imho). I guess there was a little more memory available for music on the Amiga.
Compare Lemmings, James Pond, Shadow of the Beast, or Lion King for the Amiga and the snes (on youtube or emulated).

But the Amiga had different problems. It only has 4 (hard panned) channels and 8 bit samples (instead of 8 channels and 16bit samples on the snes).

[EnergeticEssence]

How do you create an instrument? That's a really good question and I've asked myself that a lot of times. I've always wondered, how people create all those awesome sounding samples used for video games. I guess they use keyboards, and mix etc.

I have ideas in my head a lot of times, but I don't have the right samples. I guess you need to use samples (from a keyboard f.e.) and then you mix until you get, what you want. Starting with simple waveforms, and then you mix etc., until you get the right sample.

[Blitz Lunar]

It's already been explained in the thread, but there's a number of ways to get sounds:
- Sample libraries and sample CDs. This is usually the best option if you want recorded acoustic instruments, since synthesis isn't good enough for these kinds of sounds (yet!) So if you want orchestral sounds or sampled drums or guitar or ethnic instruments or vocals or whatever that sounds like "the real thing", this is usually a good bet.
- Romplers (synthesis with read-only samples.) It can either be sample-aided synthesis or synthesis-aided sample sets - basically exactly what Mark7 said. String patches for instance might mostly be samples, but the square waves and other synth sounds are likely to be shortlooped waves used with synthesis. This is what most "workstation" keyboards and sound modules use, and many "synthesizers" to a lesser extent as well. Very popular in the 90s when storage for large samples wasn't really possible. Most of the sounds you hear in SNES, N64, Playstation etc. games are sourced from romplers.
- Synthesizers. Probably the best thing to use if you're really looking to invent sounds from scratch. Of course there's many different types of synthesis (additive, subtractive, FM, phase mod, virtual modeling and so forth) that's useful for different kinds of sounds.
- Recording or sampling sounds/instruments yourself. Earthbound did this a lot. It's something you might want to fall back on if there's no other better or more practical solution for what you need.

When it comes to designing sounds, it's basically practice. It's something I'm not good at either, but you need to become intimate with how different kinds of sounds are created and better yet grab a synthesizer and attempt it yourself. I always thought the synth1 VST is a great starting point. Perhaps you could mention some specific sounds you like and I might be able to take a good guess at where they come from or how they were made.

Basically this is a huge question and I'm not even sure I just answered it properly, but it's a start.

[reversalmushroom]

I was wondering about the conversion to 16-bit. It was said that they simply lowered sample rates to meet hardware limitations, but surely there is more to it than that. What I mean is that when you listen to SNES music it doesn't sound like muddy, crappy, muffled, underwater versions of real instruments, even with interpolation turned off; instead the conversion seems to transform the samples into new and different instruments, but there isn't a real loss of quality.

[Blitz Lunar]

In most cases they aren't converted from real instruments in the first place. If they were, maybe then they would sound like ass. It depends on the soundtrack. Some sound really great (Vortex), others sounds pretty spectacularly bad (Granhistoria.)
Most SNES patches are pretty short waveforms with looppoints and ADSR settings to dictate the character of their sound. Short looped sounds tend to sound very sharp and bright.

also

... 16-bit ... sample rates ...

you're confusing two issues.

[reversalmushroom]

I was under the impression that they recorded real instruments and then lowered the sample rate and that made it 16-bit. You say that they didn't do that ("in most cases"), so where did the samples come from? How were those short waveforms made?

I was reading an interview about Super Mario Kart, and they said that they recorded real go-kart motors for the motor sound effect, and it doesn't sound like a low quality version of a real motor, and the instrument that plays from 0:45-0:56 in Super Bomberman 5's Space Station is actually someone saying "Hit it!", but that instrument doesn't sound muffled at all, so how did they convert them to 16-bit? Converting recorded sounds to 16-bit is something that I'd really like to know how to do.

When you synthesize an instrument, what determines the kind of instrument that it will be (like brass or electric guitar)?

If sample rates don't dictate how many bits an instrument is, then what does, and why do 16-bit samples take up less space than samples of real instruments? If I were trying to synthesize an instrument, how would I make it 8-bit or 16-bit or however many bits real life is?

Could you please synthesize a brass instrument and post it here? And also a non-8-bit chiptune if it isn't too much trouble?

[Blitz Lunar]

sample rate and bitdepth are two different things. 16 bit is what CD quality audio uses. it's possible that SNES samples are 8 bit (I don't actually know), but that won't make a great degree of audible difference by itself, it just means the sounds will have a higher noise floor (less space for dynamics.) I think the SNES sample rate is 32KHz, which is also only a little bit duller than CD quality. Either way, that alone isn't responsible for the sound quality of the SNES, because we're not dealing with streamed audio here, we're dealing with sequences+samples, and it's the samples that are generally speaking small since there's only a limited amount of space on the cartridges and in the music RAM (65KB at any one time/per song, unless samples are swapped in and out of RAM on the fly.)

for SNES and rompler samples generally, some parts of the sounds are recorded from real instruments or synthesizers, sometimes just the transient attacks on the samples, so that you get the character of a trumpet or a vibraphone or whatever. but then they just use that and use a short loop for the rest of the sound most of the time, plus ADSR settings. that's what creates the characteristic sound. believe me, if you knock the attack transient off a lot of these sounds, they're just going to sound like chiptunes/simple waveforms.

I was reading an interview about Super Mario Kart

wouldn't be the one conducted by Dave Harris would it?

Super Bomberman 5's Space Station is actually someone saying "Hit it!", but that instrument doesn't sound muffled at all

It isn't a very big sample. Possibly it was sampled at a low pitch and then upsampled in the song, I don't know. There's nothing particularly special or "16 bit" about it. It probably is a bit muffled compared to where it was sampled from though.

If you want to hear how muffled sounds can be, check out Tales of Phantasia's opening song, the one that can't be dumped to SPC. It has a vocal line. It's seriously muffled because the samples had to be small, likely they were sampled at high pitch and then downsampled for the song.

When you synthesize an instrument, what determines the kind of instrument that it will be (like brass or electric guitar)?

The way it's designed, the kind of harmonics it has. That's kind of a complex question really. Synthesis is a big area and there's lots of ways to get the qualities of certain instruments through additive synthesis, FM, ADSRs, LFOs etc.

If sample rates don't dictate how many bits an instrument is, then what does, and why do 16-bit samples take up less space than samples of real instruments? If I were trying to synthesize an instrument, how would I make it 8-bit or 16-bit or however many bits real life is?

as I explained, they're not 16 bit samples, but I get what you mean, and it's because there's less actual audio data than sampling an entire instrument note. short loop points.

And also a non-8-bit chiptune if it isn't too much trouble?

sorry, you're really confusing lots of different concepts. 8-bit and 16-bit when used to describe NES and SNES are describing the computing architecture of the systems themselves. it isn't to do with the audio quality. a chiptune written with 16bit 44.1KHz quality but still using a very simple subtractive synth is going to sound pretty similar to one using 8bit 22KHz.

[Solaphar]

Hi reversalmushroom, mind if I call you RM?

Anyway, RM, when you look at an audio wave form, bit depth is your resolution in the vertical, while sample rate is your resolution in the horizontal. There's a bit more too it than that, but that's the basic gist and probably the simplest explanation.


I saw a nice set of pictures for this very subject in the recent past, which had rectangles and stuff, and if I can find them, then I'll post them or a link to them (since a picture is worth a thousand words, as they say).

In the meantime, just make do with this little chart that I found at http://www.tweakheadz.com/16_vs_24_bit_audio.htm

[reversalmushroom]

About how long does it take to synthesize 1 instrument?

[Blitz Lunar]

in terms of computational time, none at all. it's instantaneous.
in terms of the person designing a sound... as long as a piece of string. you can make a very simple instrument very quickly, in under a minute or even a few seconds. or you can spend a long time working on a sound, maybe hours. depends on what you're doing, and depends on your skill/expertise as well. and what you're using.

I recommend downloading a free VSTi like synth1 and playing around with it in a VST host, might help you to understand the process behind synthesising sounds.

[reversalmushroom]

Can instruments that at least resemble real instruments be synthesized from scratch without the use of prerecorded sound samples?