Some questions about transcoding quality. by 1337haXXor at 11:03 PM EST on November 19, 2015
Long time lurker, first time poster. Long story short (as I can make it), I decided in the last year to upgrade my huge music library to lossless. Read up a LOT on sequenced music and decided to make FLAC rips converted from native chip format. I realize the difference for most video game music between FLAC and mp3 is nominal at best, but I want it for a number of reasons: archival purposes (convert later, whatever), consistency (using ALL FLAC), and most importantly, library management and metadata. As I understand, I can get a sequence player, but it won't have standard "library" aspects and you can't append metadata to it (information, playcounts, etc.).
ANYWAY, the question: I've been ripping the chips to FLAC at 16/48k, but I just read on a few posts here that the source isn't 48k, so that could cause problems? So question is, IF I'm making FLAC, how do I figure out the highest, streamed lossless quality of the source file for conversion? After reading these posts, I did some spectographing myself and noticed some VERY peculiar things (little information after a certain point, then bands of information at 16k and 22.05k, or a mirroring effect past 22.05k). What's happening?
By the way, I'm using foobar to convert from chip to FLAC.
Side note, is there any program out now, or being worked on, that can: play standard AND sequenced music, append metadata to sequenced music (if only within the program), and is available on desktop and Android? If so, that would be my DREAM and I could just use the native format, but as it stands, my library management is slightly more important.
"- Accumulator: is there an accumulator or not ? If not, what does combine operator outputs together and how are channels ouputs to the DAC ? It seems to be acknowledged now that channels are indeed multiplexed (interleaved ?) and that the "real" ouput samplerate should therefore not be VCLK/144 (~53kHz) but VCLK/24 (~320 kHz) with each 6 channels being sampled succesively. I wonder how that matters in term of emulation: doesn't that means that clipping can not occurs and that adding channel ouputs then applying a limit could be wrong ? How can we accurately emulate channel multiplexing without multiplying the output samplerate by 6 ?"
Conclusion: An absolutely correct capture of Genesis FM + PSG (mixed) output sample rate would be the lowest common multiple of about 320 kHz (FM) and the PSG chip's sample rate.
I doubt anyone would want to store an audio file that large.
Meta conclusion: The answer to your question depends on the emulated system. And it could be difficult to determine.
Meta conclusion 2: Spectographs look cool, but using them to determine the correct sample rate is impossible.
edited 5:06 PM EST November 29, 2015
by Kurausukun at 11:41 PM EST on November 29, 2015
What you were doing with 16-bit FLAC at 48kHz is your best bet. While it's POSSIBLE you might lose some information compared to the original signal, it's really not noticeable, and, as pointed out above, it would be ridiculous to try and create a file that big. I've generally found that 48kHz is a good point of diminishing returns.
That's great to hear. I guess I'll just keep at this until they release a cross platform chiptune player that can append metadata. Also, what is the best way to analyze the.. integrity of a file if Spectographs aren't the best?
That's true, but it's the only one that I know of that has a set sample rate. If you look at the actual samples, you will see that every one has a rate of 32k, but other systems are varied.
Ok, so the output rate differs by game, how would one go about finding out what a specific game's output rate is? And, I guess more importantly, what happens if I encode it to an incorrect rate (higher), just a waste of space?
Monty explains why excessive sample rates are a bad idea. 16/48k should really be good enough, if the sampling is done correctly. If you truly want to archive a source that you can derive any other format from, that's what the native emulated formats are best at.
foobar2000 with its many chip format plugins might be a good choice, I understand it does a decent job of library management and tagging, though I have little experience with it.
I don't get why you'd like to have consistency, other than that you want to include metadata into the files. I'd recommend you to not convert everything to FLAC.
Instead always keep the originals and convert to FLAC/MP3 if necessary (e.g. listening on iPod, whatever). You can archive them by also archiving the source of vgmstream, out of which you would compile a program able to convert or stream the originals.
I don't get why you'd need library management inside an application. Fileexplorers like Explorer/Finder/Nautilus etc. let you easily manage your stuff in directory structures you want. Plus, there is library management software, that compiles the list of your files and adds metadata to it's own library, just providing links to the actual files. You then put them into your favorite music player and let it handle the playlist, not the library.
Thus you keep the best quality, keep the archive for long time, and manage easily.
Great read, hcs! I know people keep going back and forth about what's actually discernibly different and what's "audiophile" audacity. I'm just looking for the happy medium.
Yeah, one of the most important things for me is metadata. I have all of my music meticulously organized and maintained, and that's the thing that kills me about the native formats: no metadata support. Certain programs (foobar2000 and others) can append metadata, but not write it to the tag, just save it as readable in the program. This is good, but if another program comes out that's better, I'd have to rewrite and append ALL the tags to my native format music, thousands of songs... That's why I was figuring FLAC as a happy medium. It's big, but I have tons of space, so it's not as much of an issue (though of course I'd LOVE to cut the space they take up by, 99+%!).
But what you're saying was pretty much my only other sufficient option, and I struggled for months debating whether to FLAC everything or do as you said, and keep native and convert/stream for on the go playback.. Now I'm revisiting it..
I guess the main reason I chose to FLAC my files is the whole switching programs thing. I'm using mediamonkey right now, which works great on my PC and Android, but THEORETICALLY, in the future, I'd like to see a program that is cross-platform (Android <-> Windows/Linux) chiptune player that appends metadata locally. But I don't want to do the hours of work to append metadata locally in a program only to switch when my "dream" program comes out, and I'm not holding my breath since I can't even imagine how few people there are out there that share this desire. The main, MAIN reason though, related to metadata, is playcounts. I use them religiously and a majority of my playlists and library is based on them. With mediamonkey, it syncs back playcounts from my Android device to my PC, and that's very important to me. I don't know that a program could sync back that type of metadata to a file with only LOCAL metadata on the server (PC).
Sidenote, Foobar2000 FINALLY announced their mobile program after years of silence, but as I see it, the only thing out right now is a Windows phone version in beta and it doesn't appear to include support for the components, but I'm trying to find out more.
So I've been talking with someone who's a part of the beta and they said that Foobar mobile is planning on adding chiptune support shortly, so that's great news! Now if the mobile and desktop version will just play nice I won't have to worry about any of this stuff anymore!