i have some questions about playing back 24bit/96kHz FLAC files from my PC through my receiver. first, i have a Pioneer VSX-D414 receiver that is advertised with "192 kHz 24 bit D/A converter". so i assumed, it can handle the 24bit/96kHz FLAC files. i have it connected to my PC's sound card (X-Fi Titanium Fatal1ty Champion series) through it's SPDIF OUT TOSLINK. in the SPDIF OUT's properties, i have it's "default format" as "2 channel, 24 bit, 96000 hz"). when i select that, my receiver displays "PCM 96" steadily. i play the FLAC files through Foobar2000 which recognizes they're 96000 hz and 3000+ kbps, but i really don't notice any difference in quality compared to having it set to 16 bit and 44100/48000 hz. so i was wondering if 1) i have the right equipment to properly play these files and 2) will the difference even be noticeable?
will the difference even be noticeable?
Unless your receiver was poorly designed or broken in some way any answer other than no would need to be backed by double-blind test results in order to be taken seriously on this forum.
i play the FLAC files through Foobar2000 which recognizes they're 96000 hz and 3000+ kbps, but i really don't notice any difference in quality compared to having it set to 16 bit and 44100/48000 hz. so i was wondering if 1) i have the right equipment to properly play these files and 2) will the difference even be noticeable?
(1) Sounds like you do have things set up right.
(2) In general there are no audible differences or benefits due to the use of audio delivery formats above 16/44. 16/44 is actually an overkill format. If you chose 12/28 as your baseline format, then 16/44 might sound somewhat different and/or a little better.
So Arnold, what do you meant by overkill? Does it mean that HD audio has little benefit compared to 16/44?
"HD audio" ??
Comparing 16/44.1 vs 24/96 is not the same as comparing DVD vs Blu-Ray
HD video generally offers a noticeable improvement over standard definition, but in audio anything above 16/44.1 is far from what your ears can reach so yes, it has little benefit, or none at all.
Well one can always make the case for cranking the volume on reverb tails or other extremely quiet passages without any regard to what happens during loud passages (assuming that the loud passages would actually reach full-scale).
Are your 24/96 FLAC audio files converted from 16/44.1 source or are those true 24/96 recordings?
Juha
Are your 24/96 FLAC audio files converted from 16/44.1 source or are those true 24/96 recordings?
That wouldn't make much of a difference.
HD Audio is a marketing hype, because the audio industry wanted something just as cool as "Full HD".
That wouldn't make much of a difference.
HD Audio is a marketing hype, because the audio industry wanted something just as cool as "Full HD".
Hmm... that's an opinion?
Juha
Yes, based on the threads here already covering this topic. But feel free to convince me of the contrary.
What about the dogs, people? Does anyone here ever think of the dogs?!
I can barely detect the negligible difference between a well-encoded MP3 @ 128 Kbps and its lossless source...but my dog surely could. I'm almost certain of it. I wish I could get ol' Heinrich to stop licking his business long enough to engage in an ABX of 16-bit CD-A and 24/96. If I ever hear that difference I'll be posting the results on this forum.
...don't hold your breath for that.
People have also started referring to blu-ray and HD-DVD audio formats (DTS-MA and Dolby TrueHD) as "HD audio" in the past few years though, even though they can still be 16/48. It's an euphemism for "lossless" in this regard.
Lossless audio in movies is a pain in the ass anyway, at least for PC playing, due to additional HDMI DRM restrictions.
Err, record a dog-whistle on 24/96 and see if Heinrich responds to the playback, to get him into training?
Edit: in reply to post #11, of course.
Juha: they are true recordings to my understanding. they're the 45rpm vinyl rips of Metallica's albums that are floating around the internet. in the text files included, they show the logs and such.
my obvious question would then be, why all the hype about this extra quality? i'm not just talking commercially but even just downloadable music. everybody rips mp3s at 320kbps, 1000kbps+ FLAC, and now a lot of these 3000kbps+ 24/96 vinyl rip FLAC's. if 16/44.1 is pretty much the max that the human ear can hear. are people like me and just like knowing they technically may have the higher of qualities?
The original is vinyl then, not 24/96 audio.
So since the original is vinyl then there's another obvious question that should be asked first. Do you think vinyl sounds equal or better than 24/96 digital audio? Cause if vinyl sounds worse, you're not making it better by ripping at 24/96.
So Arnold, what do you meant by overkill? Does it mean that HD audio has little benefit compared to 16/44?
Little or none as far as delivery formats go. There are some infrequent situations where moving up to 24 bit samples and/or higher sample rates can have some indirect audible advantages during audio production. But once the distributed recording is mixed, mastered and finished, 16/44 is all we need.
Both SACD and DVD-A basically tanked in the marketplace and failed in their bids to become mainstream. This was due to the fact that they failed to convince the general public and the scientific community that they had any sonic advantages. This contrasts with the audio CD and DVD video, which obtained unprecedented public acceptance, partially due to improved sound quality. In the case of video DVDs, the improved sound quality was due to the dramatic benefits of using more than 2 channels, and not due any improvement in the sound quality of individual channels.
It is generally conceded that the lossy compression formats that are widely used for music delivery over the web and in portable players do at least occasionally cause some change or loss of audio quality. Yet, these formats are widely accepted and have become mainstream. I take the fact that these formats are acceptable to so many people despite the fact that they sometimes audibly change or degrade music from 16/44 sources, as proof that 16/44 is for most practical purposes, an overkill format.
In the laboratory, it has been repeatedly found that down sampling commercial music recordings produced in so-called high resolution formats to 16/44 simply has no audible effects whatsoever. Thus we have reliable evidence that 16/44 is at least sufficient for music reproduction that is generally indistinguishable from so-called high(er) resolution formats.
Some of us have experimented with more extreme reductions in resolution such as 13/32 and 12/28. Depending on the musical selection, even these formats can be either indistinguishable or only barely audibly degraded as compared to 16/44. Therefore, we have additional evidence that in general, 16/44 is an overkill format.
The original is vinyl then, not 24/96 audio.
So since the original is vinyl then there's another obvious question that should be asked first. Do you think vinyl sounds equal or better than 24/96 digital audio? Cause if vinyl sounds worse, you're not making it better by ripping at 24/96.
well, since you're saying the albums are not true 24/96, i have nothing to compare to. i have compared these rips to the previous versions i had of the albums (~1,000kbps FLAC) and no i don't really notice a difference really. those compared to mp3's at 320/192/128, etc... i have not done yet. but basically, anything more than a 128kbps/16 bit/44,100 hz .mp3 file will not present any noticeable quality gains?
128kbps/16 bit/44,100 hz .mp3
Mp3 files are 32 bit float in the frequency domain; 16-bit is meaningless here. Anyway, no one is saying that 128kbps mp3 is a guarantee of adequacy, but regarding vinyl you are hard-pressed to find something that requires more than 14 bits, tops. You're also hard-pressed demonstrating that you need more than a 44.1kHz sample rate as a delivery format despite the fact that vinyl is actually capable of delivering greater than 22.05kHz unless there is something wrong with your DAC.
Unfortunately the rumour about 24bit or high resolution material is superior to 16/44.1 in every respect has spreaded all over.
On many places you can read how much better the 24bit version sounds be it a DVD rip, a HD download or somewhere else sourced from.
Mostly people compare this 24bit version to a cd version they have around from a while back. To them there is no doubt it sounds better cause of the 24bit of the new material and ignore the fact that it is for sure a complete different master transfer/mix.
It is even hard to convince people that don´t have a clue about this music being data that they compare apples to oranges then. Just lately i was in a thread with someone comparing cd vs 24bit like VHS to Blue-ray. Even if some people tried to explain he should compare the 24bit downsampled to 16bit beneeth each other and not an older cd against his new 24bit download he didn´t get it. He knows both have the same source so the sound difference is the bits - BASTA!
What can i say? In the next thread he entered again how superior his system sounds with 24bit... thats how that BS about 24bit spreads.
Very frustrating if you ask me.
Wombat. I remember that thread. That poster was quite thick headed. :-)
If you want to engage in looking for the audible benefits of 24/96 you should also make sure the equipment you use can actually reproduce sound at that spec. What I'd especially look for is SNR (of the amp) and the frequency response of the speakers.
I don't know what speakers/headphones you use, but for the receiver the SNR values are as follows:
Signal-to-Noise Ratio
(IHF, short circuited, A network)
CD, DVR/VCR, CD-R/TAPE/MD,
DVD/LD, TV/SAT ......................................... 96 dB
Signal-to Noise Ratio
[EIA, at 1 W (1 kHz)]
CD, DVR/VCR, CD-R/TAPE/MD,
DVD/LD, TV/SAT ......................................... 79 dB
(manual from here (http://audio.manualsonline.com/manuals/mfg/pioneer/vsxd414_d514.html))
I'm not sure what the difference between IHF and EIA specs is, hopefully someone can explain.
Many speakers can't reproduce sounds above 22 kHz, AFAIK. (Although I think the specs given by the manufacturers are supposed to be the only for the ~flat frequency response - so there could still be
non flat frequencies beyond those mentioned in the specs.)
Beyond having to establish the technical limitations of the playback equipment, you have to know that the producer of the recording thought there would be some benefit in reproducing frequencies more than an octave above what humans can hear. If that octave can't be accounted for during recording and post-production, then what use is it anyway, even if we could hear it?
In other words, in most cases we have real people making decisions about how to make a recording based on what their ears tell them. Their ears aren't telling them shit about what's above 20 kHz.
Unfortunately the rumour about 24bit or high resolution material is superior to 16/44.1 in every respect has spreaded all over.
On many places you can read how much better the 24bit version sounds be it a DVD rip, a HD download or somewhere else sourced from.
Like this place , for example:
www.hdtracks.com (David Chesky's new venture)
Q: Will I really hear the difference between the various formats?
A: You should hear a substantial difference when listening to the music on a home stereo. The music will sound cleaner, the bass will be tighter and you will notice a higher definition in all the instruments.
\
Is that a promise, Mr. Chesky?
I think one of the reasons this is so pervasive is that it's so easy to believe, and to understand why it's not true, you have to actually make an effort and know how things work (like the LCD monitor!), and/or understand scientific issues that can get very technical. It's about the purest form of pseudoscience. It's not like they're saying they saw Bigfoot or a UFO (at first glance, at least). I've been reading all this for years, and there is still some stuff (which usually require maths understanding) that I have to take on authority from the likes of some of you people.
I've been reading all this for years, and there is still some stuff (which usually require maths understanding) that I have to take on authority from the likes of some of you people.
This is true for many including myself, but the crucial difference is that here, you ask those authorities what's in the box* and they'll open it up for you and explain every part. And if you have questions about that, they'll open up more boxes. And so on, and so on, until it starts to make sense.
An Audiofool authoritay will simply shrug it off with, Hey, It's 24 bit!.
*) conceptually speaking, though it could also apply to a piece of hardware.
I've been reading all this for years, and there is still some stuff (which usually require maths understanding) that I have to take on authority from the likes of some of you people.
Many of the statements you see around here about "It doesn't make a difference" can be clarified if you do your own DBTs using software-based ABX comparators like the one in Foobar.
OBVIOUSLY 24 bit is better. You know, for those situations where you need your sound to sound perfectly fine, and then have something else sound fine SIXTY FOUR TIMES LOUDER right next to it (arguably up to 255 times louder).
Hypothetical question, does it make sense that a 24Bit audio interface could sound better than a 16bit one in real-time mixing environments with lots of voices (like in video games?), or is most mixing done at a better than 16bit process regardless of the DAC?
i play the FLAC files through Foobar2000 which recognizes they're 96000 hz and 3000+ kbps, but i really don't notice any difference in quality compared to having it set to 16 bit and 44100/48000 hz. so i was wondering if 1) i have the right equipment to properly play these files and 2) will the difference even be noticeable?
(1) Sounds like you do have things set up right.
(2) In general there are no audible differences or benefits due to the use of audio delivery formats above 16/44. 16/44 is actually an overkill format. If you chose 12/28 as your baseline format, then 16/44 might sound somewhat different and/or a little better.
?No way! 8bit ROCKS!!!! I don't need more than 10kHz either. Come on, get real! 16bit is overkill mate.......
i play the FLAC files through Foobar2000 which recognizes they're 96000 hz and 3000+ kbps, but i really don't notice any difference in quality compared to having it set to 16 bit and 44100/48000 hz. so i was wondering if 1) i have the right equipment to properly play these files and 2) will the difference even be noticeable?
(1) Sounds like you do have things set up right.
(2) In general there are no audible differences or benefits due to the use of audio delivery formats above 16/44. 16/44 is actually an overkill format. If you chose 12/28 as your baseline format, then 16/44 might sound somewhat different and/or a little better.
?No way! 8bit ROCKS!!!! I don't need more than 10kHz either. Come on, get real! 16bit is overkill mate.......
what the hell are you talking?
Someone is having an early christmas drink.
Unfortunately the rumour about 24bit or high resolution material is superior to 16/44.1 in every respect has spreaded all over.
On many places you can read how much better the 24bit version sounds be it a DVD rip, a HD download or somewhere else sourced from.
Like this place , for example:
www.hdtracks.com (David Chesky's new venture)
Q: Will I really hear the difference between the various formats?
A: You should hear a substantial difference when listening to the music on a home stereo. The music will sound cleaner, the bass will be tighter and you will notice a higher definition in all the instruments.
\
Is that a promise, Mr. Chesky?
I never even checked this site out before. Over-the-top "promises" aside: I for one can appreciate an online source that sells lossless. Seems like a crappy selection by my standards and preferences though. I did see some Buckethead on there that I'd consider purchasing for the reasonable price here. (http://https://www.hdtracks.com/index.php?file=catalogdetail&valbum_code=702397740926)
Is everything on here 24-bit or is some 16-bit as well? I can't tell...
I just read this crap about AIFF being superior in audio quality to FLAC. (https://www.hdtracks.com/index.php?file=staticpage&pagename=which_kind) It seems evident you just can't have a good thing without the audiophoolery to go along with it
OBVIOUSLY 24 bit is better. You know, for those situations where you need your sound to sound perfectly fine, and then have something else sound fine SIXTY FOUR TIMES LOUDER right next to it (arguably up to 255 times louder).
Obviously also, n bits are better than m with n > m, if you use the (n-m) plus bites to reduce quantisation error, but in real world applications I really don't think someone could ABX from a (24-16) bit linear quantisation noise.
I just read this crap about AIFF being superior in audio quality to FLAC. (https://www.hdtracks.com/index.php?file=staticpage&pagename=which_kind)
I don't see that stated. About FLAC, it says "FLAC files are lossless, there is no quality loss", which is a more explicit quality statement than about AIFF. The article screws up the "compression" term though -- I'm making a guess that this error comes from http://en.wikipedia.org/wiki/Audio_Interchange_File_Format (http://en.wikipedia.org/wiki/Audio_Interchange_File_Format)
Check the description of AIFF with emphasis on the latter-half:
"Even though compression shrinks the file into a smaller size, therefore allowing for faster streaming, and less space being taken up on your hard disk, compression slightly affects the sound quality of the file. Compression eliminates certain parts of songs that are outside the normal hearing range of a human being, and therefore unidentifiable by the listener. However, some people can distinguish between compressed and uncompressed files even on cheap stereo systems, while others cannot tell the difference."
There's no qualification or distinction here between lossless and lossy compression. It states that compression "slightly effects the sound quality of the file." Compare that to the description of MP3: "MP3 files are compressed therefore eliminating any sounds that are outside the normal range of a human’s hearing ability." Pretty confusing or just contradictory nonsense?
Let me correct my statement to read "this crap about AIFF being superior to compressed formats (which include lossless codecs other than FLAC.) ...and I don't even know what they're talking about anymore by the time I get down to the MP3-description."
OBVIOUSLY 24 bit is better. You know, for those situations where you need your sound to sound perfectly fine, and then have something else sound fine SIXTY FOUR TIMES LOUDER right next to it (arguably up to 255 times louder).
Not true. An obvious TOS 8 violation.
Hypothetical question, does it make sense that a 24Bit audio interface could sound better than a 16bit one in real-time mixing environments with lots of voices (like in video games?), or is most mixing done at a better than 16bit process regardless of the DAC?
Not true. Another obvious TOS 8 violation.
Hypothetical question, does it make sense that a 24Bit audio interface could sound better than a 16bit one in real-time mixing environments with lots of voices (like in video games?), or is most mixing done at a better than 16bit process regardless of the DAC?
I think this a fair and interesting question, though the wording is problematic. To reword the question, does the DAC facilitate the mixing of sounds in video games? No, it doesn't. There must be a final "mix" going into the DAC, and there doesn't seem to be evidence that output from a 24 bit input stream is going to sound different than the output from a 16 bit input stream. This is what we're saying about audio in general, so it would apply to video games.
I don't know how video game audio is mixed, but I would like to know. I would like to know what format the audio files are in. Are they lossy? Are they loaded into memory to facilitate close to realtime audio effects?
Depending on the hardware and software that one has, mixing may be a software process, or a hardware process. Soundblaster AWE32/64, Live, Audigy and X-FI have all them had the ability to mix voices in hardware.
This hardware mixing was made available to games if they used DirectX (EAX and all those standards are APIS that may be implemented in software or in hardware). Also, this hardware mixing was also offered for MIDI playback, with the combination of soundfonts.
This is why, sometimes, you can find soundcard benchmarks for videogames.
The mixing is, of course, handled by a dedicated chip, with its own characteristics. ( the EMU10Kx and EMU20Kx chips ). It can be as good or as bad as any software mixing solution.
Some soundcards needed audio to be in the soundcard's memory, while others have had the ability to access the main computer memory (Just like Graphics cards have internal memory but can use a lot of bandwidth between the main computer memory and the card). Generally, it's a compromise depending on amount of dedicated memory, bus bandwidth allocation, desired latency and dynamism of the signal (i.e. how much it changes during the game).
Said that, music is usually just streamed aside of the game sounds. And both things have used compressed (lossy) formats for a long time. Not a necessity itself, but a way to pack more audio in the disc.
Check the description of AIFF with emphasis on the latter-half:
SNIP
There's no qualification or distinction here between lossless and lossy compression. It states that compression "slightly effects the sound quality of the file." Compare that to the description of MP3: "MP3 files are compressed therefore eliminating any sounds that are outside the normal range of a human’s hearing ability." Pretty confusing or just contradictory nonsense?
True, but as I read it they're guilty of sloppy writing rather than believing in woo. I would guess that they first made the contrast between AIFF and MP3, and then remembered they had to put in FLAC, and didn't rewrite. They do better at https://www.hdtracks.com/index.php?file=sta...e=highdef_music (https://www.hdtracks.com/index.php?file=staticpage&pagename=highdef_music)
Of course, doubtless many of their customers believe in woo, so they won't work too hard to disabuse them, and I'm not sure what virtue could even be supposed to exist in releasing a 1975 live recording (Jarrett's Koeln Concert) in 24/96, but they wouldn't be the only outfit who do good work whilst having strange superstitions. Another example is MA Recordings; I have a few of their disks, and they are beautiful, but they believe in silver interconnects, and one of my disks was part of an audiophile edition that used a green carbonate substrate! Doubtless intended to save you the trouble of messing with a magic marker.
OBVIOUSLY 24 bit is better. You know, for those situations where you need your sound to sound perfectly fine, and then have something else sound fine SIXTY FOUR TIMES LOUDER right next to it (arguably up to 255 times louder).
Not true. An obvious TOS 8 violation.
This would depend on the manner in which "better" is defined.
Let me correct my statement to read "this crap about AIFF being superior to compressed formats (which include lossless codecs other than FLAC.) [...]
Well I think the appropriate correction is "this crap about what 'compressed' means".
Hypothetical question, does it make sense that a 24Bit audio interface could sound better than a 16bit one in real-time mixing environments with lots of voices (like in video games?), or is most mixing done at a better than 16bit process regardless of the DAC?
Not true. Another obvious TOS 8 violation.
Is it 'obvious' -- or even anything close to true -- that a
question (phrased as such, not as a rhetorical question, but as a "Hypothetical question") is what the TOS refers to as a "statement"?
Is everything on here 24-bit or is some 16-bit as well? I can't tell...
It's mostly lossless Redbook, but they also offer some in 24/96 -- for tighter bass!
OBVIOUSLY 24 bit is better. You know, for those situations where you need your sound to sound perfectly fine, and then have something else sound fine SIXTY FOUR TIMES LOUDER right next to it (arguably up to 255 times louder).
Not true. An obvious TOS 8 violation.
An obvious disrespect for picking up internet sarcasm set off by noobish language, over-emphatic capitalization and describing situations that never occur (especially in a loudness war driven marketplace) and aren't even that helpful anyways. Please cool your jets and don't overreact to jokes meant to make a bit of humor out of this topic that's been beaten to death over and over.
Hypothetical question, does it make sense that a 24Bit audio interface could sound better than a 16bit one in real-time mixing environments with lots of voices (like in video games?), or is most mixing done at a better than 16bit process regardless of the DAC?
Not true. Another obvious TOS 8 violation.
I don't see how a question can possibly even be a TOS 8 violation, have you read the TOS lately? I double checked, as I try to adhere to them that maybe I overlooked something. In fact, the only thing that TOS 8 could apply to here is a response to my question that's very direct and un- or improperly supported. Please try to think of helpful responses instead of jumping to blatantly invalid claims of TOS violations.
Hypothetical question, does it make sense that a 24Bit audio interface could sound better than a 16bit one in real-time mixing environments with lots of voices (like in video games?), or is most mixing done at a better than 16bit process regardless of the DAC?
I think this a fair and interesting question, though the wording is problematic. To reword the question, does the DAC facilitate the mixing of sounds in video games? No, it doesn't. There must be a final "mix" going into the DAC, and there doesn't seem to be evidence that output from a 24 bit input stream is going to sound different than the output from a 16 bit input stream. This is what we're saying about audio in general, so it would apply to video games.
I don't know how video game audio is mixed, but I would like to know. I would like to know what format the audio files are in. Are they lossy? Are they loaded into memory to facilitate close to realtime audio effects?
Perhaps my wording was a bit off, I'll try to correct that to clarify the issue. This is of interest to me because I'm about to get a BSE in Computer Science and go into the Game Development field. I don't intend to contest any difference between sending a 24-bit mixed signal to the DAC and a 24-bit mixed signal downsampled(? not sure on the right term for this) to 16-bit sent to the same DAC. What I was mostly interested in is that if using a 16-bit buffer as the mixing target has ever been shown to be worse than using a 24-bit one when mixing many samples. It's my understanding that these mixings are done in place one at a time, accumulated one at a time (pre-dividing the amplitude or worse). If you had a large number of voices being mixed (say, 64 to 127), this represents a a pre-division factor of 7 bits out of your 16, meaning all samples get downsampled(?) to 11 bits at most, assuming there's no HDR processing like is new in visuals. Then you add a ton of these lower depthed samples together to get your completed 16-bit buffer. My question is basically two part: 1) is how this mixing is done fundamentally different from how I guessed thus avoiding any potential problems? or 2) Is there any evidence that indicates that using a 24-bit buffer to mix into sounds any better (since all the signals should still have a pretty wide bit depth after pre-division)?
To answer answer your question about formats: game sound effects are frequently lossless, but don't have to be. More than a decade ago it wouldn't be too uncommon for games to have completely lossless sound tracks, sometimes even being written to the disk as an audio cd playable in a cd player alongside the game. This doesn't happen anymore, music is almost always compressed because games frequently have huge OSTs these days. As for the formats: if a game uses DirectX for its audio, it likely uses ADPCM or a computationally cheaper version of WMA (xWMA). If the game uses OpenAL for its audio, it most likely uses vorbis audio files. MP3 is common in a lot of systems too as DirectX, OpenAL, and many of the other lesser used platforms either support it directly or allow custom format importing. If you're looking at a console game, each console has only a few types of audio codecs it can give hardware support for decoding and games tend to use those.
Hypothetical question, does it make sense that a 24Bit audio interface could sound better than a 16bit one in real-time mixing environments with lots of voices (like in video games?), or is most mixing done at a better than 16bit process regardless of the DAC?
There's nothing magic or special about mixing a lot of voices. Consider a choir.
All sounds are composed of the summation of a large number of enveloped sine waves. It is possible to determine which combinations of waves are the most difficult to process without adding soloration and spurious responses. Once you have a system that processes the most difficult situations well, the rest is well, easier.
I think he meant about noise adding up like when you do lower bit depth consecutive edits.
What I was mostly interested in is that if using a 16-bit buffer as the mixing target has ever been shown to be worse than using a 24-bit one when mixing many samples.
I believe that I mentioned in an earlier post that > 16 bit samples and > 16 bit processing have and commonly obtain a legitimate place in certain audio production processes. Mixing is one of them. Mixing is often done using floating point arithmetic that commonly involves a 24 bit mantissa. Fixed point mixing usually involves 24 bit integers that are accumulated in 48 bit accumulators.
It's my understanding that these mixings are done in place one at a time, accumulated one at a time (pre-dividing the amplitude or worse). If you had a large number of voices being mixed (say, 64 to 127), this represents a a pre-division factor of 7 bits out of your 16, meaning all samples get downsampled(?) to 11 bits at most, assuming there's no HDR processing like is new in visuals. Then you add a ton of these lower depthed samples together to get your completed 16-bit buffer.
First off, each one of your pre-divisions should be properly dithered. Whether you dither the pre-divisions or not, Your problem is that 7 bit audio has approximately 42 dB dynamic range. Assuming that the sources are uncorrelated and the dithering is uncorrelated then the sum of 2 sources with 42 dB dynamic range is a signal that has only 42 dB dynamic range. Both the signals and the noises add geometrically so their ratio remains the same. If there is an audible problem, then the problem would be a poor noise floor. But the psychoacoustic impact of that noise floor is not obvious for me to analyze.
I think that you ought to do an experiment using real world voices, etc. Do it once with a 16 bit accumulator, and do it again with a far larger (e.g. 24 of 32 bit) accumulator. Matching levels and time-sync hing the files, do your own ABX tests. If you are competent with the tools that are used to do this kind of production, you should be able to get a result within a day or less.
You seem to be unsure about exactly what is happening with the software that you are using today. You should clear that up. Speculation based on speculation is garbage.
My question is basically two part: 1) is how this mixing is done fundamentally different from how I guessed thus avoiding any potential problems?
You just said an evil word - guessed. If you don't know how to ask a relevant question you should first sort that out.
or 2) Is there any evidence that indicates that using a 24-bit buffer to mix into sounds any better (since all the signals should still have a pretty wide bit depth after pre-division)?
Most of the software and hardware tools for mixing that I am aware of use either 32 bit floating point or 32 bit (or larger) fixed point buffers to mix in. You do know that the conputer accumulators that are used to do math with 16 bit operands are generally 32 bits long, right?