Comment by pea | Hacker News Reader

Comment by pea | Hacker News Reader

Parent Full thread

What really flummoxes me is how my Yamaha SY99, which is a digital FM synth from 1991 with 512kb of memory and, by modern standards, a tiny, tiny CPU, manages to do things that are really CPU intensive on a normal OS (like large plate reverb, tons of polyphony, on 12 tracks at a time) without ever lagging. The only real limit I've hit on it is loading in proper samples, versus waveforms, though this[1] insane project seems to have hacked memory expansion boards for the old SY/TG random. This is a piece of purely digital equipment that is now 27 years old, and yet my phone doesn't last for more than 2 years without feeling obsolete.

Why is this? I've always wondered if these golden-age hardware FM synths provide any benefits over software (people talk about the super HQ DACs, etc., but I'm not sure if I buy that these components are better than modern commodity equipment?). The thing is, to my ear, the SY99/SY77 etc. do sound a lot better than e.g. Dexed or FM8, but is that just because Yamaha patented a bunch of the FM algos and architectures, and, with a retail price of £3,000 in 1991, they put more care into a complete end product than a VST FM developer realistically would? Or, can really low memory, low CPU, super purpose-built hardware somehow win against a modern OS?

[1] https://www.sector101.co.uk/waveblade.html

Modern systems largely sacrifice consistent latency for more throughput and more energy efficiency. This leads to modern systems being able to do much more, but also having frequent little responsiveness hiccups and occasional larger hiccups. If you want to do a smaller job consistently, a simpler embedded system architecture is still the way to go. This is especially true when doing any kind of low-level signal generation, whether it be audio, video, or control.

As for the sound, I don't think Yamaha's implementation of "FM" [1] has been comprehensively reverse-engineered in the way that, say, the Commodore SID has. There are a lot of little quirks and edge cases to take into account when considering the full operational ranges of the various chips. Even MAME's implementation is allegedly distinguishable even after 20 years of tweaking and testing against hundreds of games, and I imagine most VSTs are using significantly less mature code.

[1] I recall reading that the actual implementation is phase modulation because directly doing FM in the digital domain would put quantization error in the frequency domain, i.e. notes would be off-pitch instead of having noise.

Comparing difficulty between FM and SID is a bit apples to oranges. The SID has analog variance in its filter, and no two are exactly the same, so a pure DSP implementation has to do rather heavy math to approximate the filter. As a result most people lean on a handful of SID emulation cores because otherwise their emulation is just wrong.

OTOH the things that tend to differentiate implementations of Yamaha's FM are the sample rates, bit depths, and envelopes used, plus any output distortion(YM2612 for example has a well known distortion in its implementation that adds a harsher edge). The core algorithms they use are something a high school student could pick up and do something with, and emulation quality issues are more a matter of it being easy to write an emulation that gets it 98% correct without covering the last steps, since those are details that really need error-for-error reproduction of the original ASICs and boards, including any timing issues - things which frustrate emulator writers everywhere.

> manages to do things that are really CPU intensive on a normal OS

Dedicated circuits.

This isn't all in CPU - the CPU on these synths mainly sets parameters for the sound generating chips.

not the same chips, but something like this probably:

https://en.wikipedia.org/wiki/Sound_chip#Yamaha_2

Phase modulation synthesis as Yamaha does it is computationally very cheap. That was one of the things that really boosted its fortunes in the 80s: you could produce very complex timbres (albeit often nasal or harsh sounding) with just a few sine waves pushing each other around. It's literally just a sine wave whose output pushes the phase of the next sine wave (operator in Yamaha's parlance) forward. You don't even need floating point math, and Yamaha got by with 12-bit DACs.

The SY99/77 really are great, though. I sold my TG77 and FS1R recently, but will probably get myself a used SY99 keyboard at some point. Or save up for a Montage.

>

The best-known Yamaha FM synth chips don't even have analog output; they pair with a separate DAC chip.

Reply on news.ycombinator.com