Truth is, as someone who actually owns the real hardware and the necessary equipment to play NSFs on it (and has done so), I can say the hiss is nowhere near as stringent and unpleasant as it sounds in Famitracker. The tracker makes it seem way worse than it really is.
If you don't believe me, listen to this lossless recording of Show Me A Moose, versus the FTM. You'll find the issue is much more apparent in Famitracker.
Point being, either make it more accurate or give me an option not to have my eardrums pierced whenever I want to use the tracker however I please (not the hardware, the tracker).
Oh and don't give me the "I don't see why you'd want to use it if you don't care about hardware accuracy" argument. Just because you can't understand why I want to use the tracker in a certain way doesn't mean I'm wrong and I shouldn't be using it that way.
...but you can already use it however you please; it's open source, and rainwarrior already gave everyone a hint (sounds like the registry fix has been commented out, but the code is still there).
If jsr was really worried about people creating their own derivative versions, I don't think the source would be available.
If the emulation is incorrect, then yeah, I'm with you, but listening to these it's tough to say.
Take the aliasing challenge. :P I ran both versions through an 8k highpass filter (brace your ears). Which is which?
_______________________
The only things certain in life are death and uncertainty.
The tracker is a tool for listening to music, but it is also (and primarily) a tool for making music. As such it has a lot of features that makes making music easier, like the ability to mute channels, or customize other playback settings, or export individual channels for mixing, etc. Just because the target goal is hardware-accurate listening doesn't mean that all stages of the workflow also must have hardware-accurate listening.
I think it's kind of absurd to support the creation of 8-channel N163 music, but at the same time make it unusable. It's a bit of a tease, no?
Sure there's a workaround to turn on a strong highpass filter, but nobody really wants to listen to other tracks like that. What this creates is an annoying workflow, where you need to fiddle with your config settings every time you change songs.
And... I guess that's what it comes down to. Removing this feature creates an annoyance for people that use your tracker. Now the people who want to use 8-channel N163 have to mess with the filter settings a lot, or recompile Famitracker to get at the feature they already had.
As I said, I don't personally want to make 8-channel N163 music, but I want to make the pragmatic point that many people DO want to do this, and will continue to. For these people, you get one of the following results:
1. Disabled a feature which made it easier for them to make this music.
2. Driven them to start using an alternative FamiTracker build, which is bad from a testing perspective-- now this version of Famitracker is unsupported, and any bugs (which are probably relevant to the mainline Famitracker build) will be harder to reproduce and debug.
I don't think doing either of these things is a good idea.
As an aside, it's also an annoyance for people like me, who like to open FTM that others share and listen to them. When one of them uses an 8-channel N163, I get blasted with the noise, and immediately close the program. So, consider this is not just an annoyance for composers, but listeners too.
rainwarrior, I don't know if I've ever said this to you, but I admire your prose. You've condensed everything I thought about this whole issue into a single, comprehensive and digestible post. Hats off to you, my friend.
Play this ftm back at different sample rates. The multiplexing frequency is there, yes, but there's another lower frequency which is much more audible and varies according to Famitracker's sample rate. There's no way this can be accurate emulation, right?
edit: At 44100 and 48000 sample rates, I think it's the multiplexing frequency divided by powers of 2. Still sounds different between the two sample rates (I think many will find that 48000 is less annoying). At 96000 it's a different frequency.
edit: Another interesting discovery. "20 hz -0 dB" low-pass setting filters the N163 without affecting the 2A03. At all.
Accurately emulating high frequency signals is kinda tricky. On the 2A03 this problem mostly manifests on the noise channel, where subtle inaccuracies can be kinda difficult to distinguish from the real thing. With steady tones though, it can be a lot more apparent.
I've been working on N163 emulation, myself, and I've been trying different techniques to deal with inaccurate aliasing/sidebands, but it's kind of an ongoing adventure trying to get it clean. I think I'm going to implement a 1.8MHz oversampling option, so that I can emulate without skipping any CPU clocks, but I'm still trying stuff for the "normal" playback modes.
It's actually even difficult to deal with this issue with hardware recordings, since you can get sometimes aliasing/sideband leaks from your recording equipment too.
Personally I would do one single thing to satisfy both sides.
Multiplexer should only be on for normal mode, so everyone would be able to actually CREATE the music without having to listen to the multiplexer noise when we lay down notes and preview the track. And when we switch to Normal mode, we can take a listen to the hardware accurate version and make the necessary changes.
Most composers back in the day created their music without ever listening to it on the console itself, so the N163 composers never had to bear a 15k tone destroying their ears during production.
Sure, you'd say that noone would listen to anything N163 in normal mode then, but people can already cheat in numerous ways if they want to.
N163 composers never had to bear a 15k tone destroying their ears during production.
There was only one 8-channel N163 composer, Hirohiko Takayama. From what I've gathered from interviews, I think his method was to compose on midi synthesizers, and then when finished entered the notes into the game by hand as hex.
Maybe we can add a separate treble filter for the N163 only, so we can save our ears and have hardware accuracy at the same time?
Yeah, but how about an actual solution: such as using less channels or another chip?
jsr, you're being extremely disrespectful. Just because you don't agree with something doesn't mean it's not an option. Try to see the issue from the other person's side.
It'd be interesting to hear a hardware recording of multiplexing isolated from N163 output (demonstrated in the above FTM by playing a note below 20 hz). Surely a reasonably accurate recording is possible, by raising the sample rate and/or filtering the audio before it's converted to digital.
At least the hiss can be filtered in Famitracker, but aliasing cannot. If it comes hand-in-hand with emulating the multiplexing, then perhaps multiplexing should not be emulated until it is fixed?
I did once play a sub-20hz tone on my N163 cart (those of you who were in the IRC channel at the time may remember the "youth repellent" incident... good times!). It's pretty unnerving and annoying as you would expect, but playing the FTM rather than the NSF would've been more effective at keeping those annoying teenagers off my lawn, really.
That's pretty cool.
A bit of Multiplexing a day, keeps the teenagers away.
They should also use the N163 as a tool during plice interrogations, and everyone would confess kissing their cat's whatever.
