iYM2151

Imagine with Microcabin would have done with the YM2151!

First time I hear this thing, the bass is nice and the stereo too, the rest not anymore.

You may be surprised by the fact that extra operators and algorithms, the specific way they're implemented in the YM2151 (but also in e.g. the Moonsound), offer relatively only few handy new bits for a composer. Added value from extra operators will have value if each operator has fine tune and when there are multiple feedbacks in an algo. Extra waveforms, and complex signal flow (such as [4]->[3]->[2]->[1]), usually tend to complicate waveforms into a complex tone that doesn't add much compared to simple 2-op models like in the FM-PAC. Again, listen to Micro Cabin..

The best way to make an enhance FM-chip would've been to add polyphony (18 2-op FM channels is far more practical than 9 4-op channels), and add a few DSP effects such as chorus, delay and reverb and offer send parameters per polyphonic channel. Technically this would really not be very complex, the implementation techniques for such effects are rather simple in fact.

With these 18 channels I wasn't referring to anything specific. FM is just FM, whether it's a Yamaha synth, a VST or an FM-PAC.

Of course there is a difference between 2-op and 4-op, the question is whether this difference is relevant and whether it's relevant enough to offer a polyphonic channel for that. With sine operators (like a Yamaha DX7), I could see a number of applications for 3- and 4-op serial algorithms (e.g. 4->3->2->1), but since this YM2151 and the Moonsound have operator waveforms, the advantage is less relevant- though not completely zero. In my own FM8 videos, many sounds are in fact composites of 2-op algorithms. So even when I use all 6 operators, then it'd often still be 2-op + 2-op + 2-op.

So, is there a difference between 2x 2-op FM and 1x 4-op FM if this is the algo: (4->3->output, [FB]>2->1->output, known for being used as the DX/EPiano). Suredo!

With two 2-op channels you gain the following:
- a feedback per operator stack, rather than only one
- detuning per operator stack
- stereo panning per operator stack

Next to all that: basic sounds are very practical.. be it sawtooths, blocks, it doesn't matter. There's more often than not barely any point in extremely complex sounds. They may sound funny 'n cool in the voice editor, but in a musical context they're often not needed - a perfect analogy with weird preset sounds in synthesizers that sound nice in the shop but are very impractical in a musical context.

Is looking nice
The most of you will recognized this song:

http://www.youtube.com/watch?v=oX-INdxV-So&feature=related

anybody knows why there has not been any game or demo which supports YM2151 (CX5M/SFG01,05) in MSX? Is this due to some technical problem? For instance, do playing YM2151 music require too much CPU or memory resource to run normal game or graphics demo in CX5M?

There is an Demo Tape maybe I 'm willing to take it with me to Hoorn this Saturday

as much as I love msx music I dissagree. Purely judging from youtube (since I dont have a real x68000) I feel in general ym2151 sounds "fuller" compared to msx music.

for example the sharp x68000 had a ym2151
http://www.youtube.com/watch?v=OyfcUaszGmU&feature=related xak on x68000.. this also answers manuels question what microcabin could have done with a ym2151
http://www.youtube.com/watch?v=uhPSonNiMJk YsIII
http://www.youtube.com/watch?v=3cJtaN5hHtQ&feature=related Genocide 2, I really love this soundtrack :D

okay ofcourse the x68k also has 1 channel pcm, probably used for the drums so you should ignore the drums if you want to compare.

and besides, I would say stereo alone is a big step forward/improvement

There's no law that dictates how complex FM should be in order to be labelled 'FM'. 2-op FM is FM. Heck, one operator with feedback is FM already, just as having a very fast vibrato connected to an oscillator. There's no point in going complex, only for sake of being complex, the only thing that counts is the result. Now, a serial algo of 3 or 4 operators has its uses if your operators only have sines, like in the DX7. Because with that, you can make modulator envelopes making the sound go from "complex to complex", rather than "complex to sine". But as soon as an operator has operator waveforms (FM-PAC, Moonsound, YM2151), having a non-sine as carrier already means an envelope from "complex to complex". Anyway, all this is pure academic blah. Imagination and ears, that's what you need to create FM sounds, not what the books and the specifications say.

Btw:
fb[1]->[2]->out
[3]->[4]->out
is a rather standard algo in FM synths (the DX7 ePiano). 'Additive' as you call it, although true additive would refer to Fourier synthesis, which is not the case here. So, if that's already standard:
fb[1]->[2]->out
[3]->[4]->out
[5]->[6]->out
shouldn't be that alien, no?

As for which algos those synths had or didn't have: these algos often had no point, other than being a variation to the other algos. The number of people who could create decent FM sounds out of a DX7 was frustratingly low - hence, you hear the same presets being used on all these albums from the 80'. With such little user feedback for Yamaha, they probably wouldn't know what algos to make anyway. And why would users want a 3x 2-op structure anyway.. for pads? Go for a Jupiter, Synthex or Prophet, not a Yamaha DX.. or heck, a DX connected to a chorus effect. The Yamaha SY77 would later have custom FM routing, multiple feedbacks, and - more importantly - filters. Especially the latter was like finally giving in to the fact that at some point you're going to need some filters - FM synths omitted those for years.

As for stereo: yes and no. Hard-panning with a low number of polyphony is tricky business as it creates great holes in the stereo image. If you pan a sound full-left, then in the center and right you have nothing, this creates this hole. Of course you could pan another sound full-right to balance the left sound, but you get the same issue again. In a way you are making the stereo image somewhat weak if autonomic voices are hard-panned. Of course you can create a double voice; one instrument with a left sound and a (different) right sound, but that'll cost you two channels from a chips that hasn't got much of those anyway. In the Moonsound it works a bit better because you can have up to 18 channels (2-op), so there's a lot more space to counter-balance hard-panned instruments. The best way to create a stereo image would've been by adding DSP effects (reverb, stereo delay, chorus etc.), which really aren't that hard to make - it costs you some memory though, to store all the audio from the delay lines.

X68000 looks like a very nicely specced machine. Kind of like the Japanese Amiga Great thing, it seems.