Put me on the list of people who can't hear the difference between 44.1 and 96... ...I mean really. I'm not gonna get into the science behind it too much. I'm just not interested in understanding something I personally don't perceive. I've had mixed results forcing sample rates. Either same or worse, really. If there's an auto option, leave it be... ...best sample rate is the one the source is at. If not, 44 or 48. Up-sample artifacting is not nice. Some dacs and sources really hate to have that stuff messed with. I can tell you this, you're not going to get more bass than is already there at 44.1. Nor is there likely to be an increase in clarity. All you are really doing is increasing the range of ultra-high-frequency content beyond what you can hear. If you drop down to 22, you'll lose a lot of the high frequencies, as its trunicating the highs down into what you can hear. Once you hit 44.1, the entire reproducible frequency range is there.
Going higher is like having a monitor that outputs infrared or ultraviolet. Does that make sense?
Getting into a gray area, there is a theory that DACs benefit from oversampling. There is a measurable artifacting phenomenon that occurs at 44.1 that has to do with the calculations that the DAC has to do for its conversion. It's mathematically not bit-perfect by nature, meaning it's not 100% accurate. At a certain point in the process, a peice of info is lost and there's no way around it. Increasing sample rates artificially is basically supposed to drag out those artifacts that bring out measurable inaccuracies in the output completely up and out of the "hearable" range. So it's still inaccurate... ...but mostly only to dogs. And for maximum benefit the source has to be a higher sample rate to begin with. Only some dacs can do oversamling properly... ...the architecture has to be designed for it. Just because a DAC or sound card can work with high sample rates doesn't mean it can upsample properly. Most sound cards probably can't, and forcing them to will likely sound worse than the native sample rate of 44.1. Scaling it up past that can cause problems of its own.
Personally I'm not a big fan of theoretical audiophilia, or strict measurement adherence. There is a measurable difference in ultra-high sample rates, with the right methods and the right equipment. But if there's anything these past several years in the world of high end audio has taught me, it's that measurements and numbers aren't everything. I have heard stuff that measures near-perfect frequency response and nearly zero distortion that sounds like stepped-on shit, and then there's stuff that measures over 10% THD that actually sounds incredibly even, dynamic, clear, and nuanced. And sometimes the opposite is true. But there are multiple ways to look at and interpret data like this. At the end of the day, when something sounds wrong, measurements will often tell you why - for instance, ringing distortion in certain ranges with specific characteristics in the time domain will denote certain audible defects in a transducer's sound, such as "cupping" or sibilance. You can see that without even hearing it and just know.
So sometimes we can see "bad." But nobody has figured out how to measure good. A lot of measurable differences that you can see on a chart or graph, can't likely be heard, or maybe don't manifest in the ways you expect. Sometimes things that measure as less distorted or more technically accurate do not sound better. We are not even close to understanding the hows and whys of what we experience vs how the input is processed and what our speakers put out. It can be quite the rabbit hole.
Bout all I can say there. Play with it and see if you can hear the difference. If not, then there you have it. If so, there it is. Trust your ears. There are situations where I can see it mattering for playback, but past having the full audible frequency spectrum or perhaps a little more, the real-world implications are limited, outside of like, ridiculously expensive super-granular, hyper-resolving setups. We're talking several thousand bucks minimum for entry. And it may not even be worth it, not for the accuracy anyway... ...though maybe for other things. The amount of accuracy gained is so small when most dacs are already approaching 100%. 44.1 will be fine for most. The majority of your source material is mixed down to that anyway, though it may have been mastered at a higher sample rate.
Which brings me to the next point. One use case for these higher bitrates and sample rates that I CAN argue for is in the actual production phase. There are actual benefits to recording/processing tracks at 24 or 32 bit at 192 or whatever crazy high sample rate you can squeeze out.
When it comes to digital processing and mixing, headroom is critical, and upping the bitrate helps bring the noise floor down and improve dynamic range. In a sense, it gives you a finer canvas... ...a wider range of values, which is important in the digital world. It's like working with a 20-megapixel image vs a 4-megapixel image. Things will be smoother when scaled down to the final resolution for the 20mp. It makes a big difference when you have noisy, high gain sounds, or lots of layers with different dynamic signatures, especially when you're messing with dynamics a lot. I'm really oversimplifying and there's a lot more to it, but that's the gist. Mixes done at 24 or 32 bits sound better, even when mixed down to 44.1.
Super-high sample rates mostly come into play with plugins... ...circuit emulators, synths, eq's, IR's, reverbs. A lot of these things benefit from being allowed to work with frequencies far beyond what you hear. Many actually have an "oversampling" feature that jacks up the sample rate for the plugin. In these cases, what happens past the audible range actually influences the output on the audible spectrum. It completely alters their behavior. Many plugins just perform better at higher sample rates that makes sense when it comes to audible frequency range. Again, it comes down to how they calculate the output, and certain quirks of how they work causing nasty things to happen past a certain point of the total frequency range available to them. The closer the ceiling is to the audible range, the harsher they tend to sound. So upsampling works similar to antialiasing in this case... ...smoothing over nastiness to an imperceptible level.
Another reason for higher frequency bandwidth is when you're doing a lot of time/pitch altering. There's simply more information to extrapolate from... like stretching an 8k image on one axis before shrinking it down to 4k, versus stretching a 4k source image. Which will look better?
But again, final mixdown is gonna be 16-bit, 44.1. By the time the tracks are processed and bounced down to a single stereo track, the benefits of higher sampling/bitrates have been and gone. For better or worse, that's standard practice. The common consensus is that nothing important is really lost.
But hey, don't take my word for it. Seriously, just try different sample rates and see how it goes. There is so much to this stuff, I wouldn't trust anyone who claims to know the absolute truth to it all.
