How to View if Music File is 8-bit, 16-bit, etc. by MoldyPond at 11:34 PM EDT on May 13, 2019
I know where to view a music file's sample/bit rate, but I can't seem to find a way to view if it's 8-bit, 16-bit, etc., other than vgmstream telling me most Nintendo music is 4-bit in the control panel at the bottom of foobar. Does anyone know anything about this? Any help would be greatly appreciated. Please and thanks :)
Before I being, I'd like to say that my knowledge on this subject is decent, but not ideal. I know what I need to know to get by (and a little extra here and there), so I'm probably not the best person to explain what's going on here. Still, I'll give it a shot and hopefully shed some light on things, but if anyone else comes along and corrects me or even says I'm entirely wrong - then it's probably true.
First up, the report you're getting - 4-bit - sounds correct to me. This isn't a reading referring to the characteristics of the music (ie. what you can hear), but rather the encoding of the file.
There are numerous ways to store digital audio. Most people are used to seeing PCM (Pulse Code Modulation), which stores a digital representation of an analog waveform in a linear fashion, but variants such as ADPCM (Adaptive Delta Pulse Code Modulation) store data differently. Instead of storing the data in a linear 8/16/24-bit stream, ADPCM stores only the difference between the previous and next sample. Because of this, it can represent the data in a similar way using only a 4-bit stream. This isn't without drawbacks though, but that's not what I'm covering here.
The important thing to note is that these digital encoding types can represent any type of audible information. This could be an solo instrumentalist, a choir, a band - or a recording of an existing soundtrack. This means that the sound-style which is commonly associated with consoles of the 8-bit era could be represented in an compressed mp3 stream, a 4-bit ADPCM stream or a 24-bit stream. It's at the encoder's discretion and usually the format is chosen based on constraints such as disk space and tools available.
It's worth noting that an 8-bit source can't always be represented by an 8-bit PCM. There are a number of other factors which could come into play, such as the capture method when dealing with analog sound sources (which are being recorded to a digital format), or additional modifications or changes that could have taken place prior to encoding. This means that although the sound can be identified as 8-bit in sound style by our ears, it might not be able to be represented in an 8-bit PCM stream without losing some potentially significant sound information.
4-bit is a very typical reading for ADPCM and I see it quite often. If you have an example file I'd be happy to look at it and check for a false positive, but it's very likely that it's just a 4-bit ADPCM.
I hope some of that is useful, I phrased that all correctly and I didn't misinterpret your question.
Well damn, I thought the answer would be a lot less complicated than that :D
I'm trying to find out if there is any real difference between the PC, PS4 and (recently uploaded) Switch versions of Snake Pass's music, since each version is a different file type and bit rate (Vorbis on PC with lowest bitrate, ATRAC9 on PS4 with higher bitrate, and DSP on Switch with highest bitrate).
I had already asked a few months back about if there was a difference between the PC and PS4 versions because of the bitrates and found that even though the PS4 is higher, the spectograph says they're literally the exact same quality, so now it just comes down to finding out if they're all 4 or 8-bit.
Here are the files for one song from each version if you'd like to check for me, also, thanks a bunch for the response :) :) :)
The 4/8/16/32-bit thing refers to bits per sample, also known as bit depth. It describes the number of possible values for the quantized voltage signal in PCM audio. This is different from bit rate, which is usually measured in kilobits per second and just measures the raw data rate of the audio, PCM or otherwise.
Vorbis and ATRAC9 are lossy codecs (i.e., not PCM) so their quality is better described by their bit rate, and also depends on the encoder settings. They don't quantize the signal in the time domain like PCM, but do a frequency transform on the input audio and quantize and compress the frequency domain signal. They both probably get decoded to 16 bit PCM.
So basically, the only useful way to tell which one sounds better is to just listen to them. Sometimes, ADPCM will perform better on transient-heavy audio because it doesn't introduce windowing artifacts that cause things like pre-echo (a type of frequency domain quantization error). But ADPCM can also sound "bitcrushed" because the effective bit depth has been decreased (which results in time domain quantization errors).
I find that when things are close, it's best not to over-complicate the situation. The one that sounds best is the one that sounds best. If you can't tell (even after several listens), then it really shouldn't matter. I do have a hard time convincing myself this and often have to reiterate that point mentally once or twice to force myself into a decision.
That said, there are a few ways to analyze audio files and compare what's present. A quick and easy way to do this is to have a look at the spectral frequency graph (I use Adobe Audition, but there would likely be a free alternative). For the sake of this thread, I took some screenshots. I used linear view (typically I use logarithmic) to make the changes more visible in this case. This means that the visual differences aren't as significant as they appear, keep in mind that the higher frequencies are much less audible.
As you can see, there is quite a large difference between the files. Typical compression methods mask points in the track we're less likely to notice to help reduce size and this becomes very visible, acting almost like a low-pass filter (the higher frequencies are cut off). In this case the assumption that the larger files are better quality seems to be correct.
This isn't 100% reliable though. Sometimes two tracks are mastered differently, some compression and processing methods (in particular, certain resampling methods) introduce artifacts that appear to be useful audio information, despite the track being reduced in quality - and so it's important to consider this when analyzing a track.
All of that aside, it looks like the Switch copy is both highest in bitrate and has the most unique audio information present and so I'd consider it to be the best in this case.
As said above just listen and compare. If you want to "guess" the quality, codecs+bitrate and sample rate is more useful.
First, the higher sample rate the better (*very* important, 48000/44100 is a sweet spot).
Then codecs, very roughly and arguably, from better to worse: - 24/16-bit PCM / FLAC - "MDCT" codecs using high bitrate - DSP/XA/VAG/etc 4-bit ADPCM - IMA 4-bit ADPCM - "MDCT" codecs using low bitrate - 4/8-bit PCM
What is high and low bitrate depends on the codec. Note that ADPCM bitrate is basically fixed, so being higher/lower than other codecs won't tell you much.
"MDCT" codecs are ATRAC9/XMA/Vorbis(Ogg)/OPUS/AAC/MP3/etc. They don't use 4/8/16 bits per sample as you understand them, so that value is useless to compare them, and that's why you won't find it anywhere.
Also quality depends a lot on the encoder (program used to convert from PCM .wav to other codec), and possibly the song, specially for MDCT codecs. For example Nintendo's OPUS encoder seems a lot wonkier that the latest official OPUS encoder.