The choice often comes done to theory versus practical. If a 16/44 DAC does it job perfectly and the brick wall filter is irrelevant, then 16/24 should perfectly reproduce the original sound. However, DACs are not perfect and brick wall filters do matter. The issue is how much they matter. Some DACs sound better with higher sample rates, some do not. It depends on the implementation of the DAC. In fact, 16/44 files externally converted to 24/192 can sound different than 16/44 files fed to the same DAC that internally converts to 24/192.
As to the "adding zeros' comment. I see that a lot. Converting for 16 bit to 24 does add a bunch of zeros to the end of each sample. However, converting 44 to 176 or 192 definitely does more than add zeros. It does an interpolation between the original data points to come up with the additional points. It is not a normal linear interpolation, but rather the surrounding data is fit to a mathematical function that takes into account the local pattern and determines the intermediate points from that function. Sounds complicated but the various algorithms are pretty well known. It just depends on which ones the particular converter uses.
As to JRiver, there are people on their forum asking for different upsampling converters, because they believe there is a difference in them. It can be a lively discussion, because there is far from a consensus on this. The JRIver principles seem to believe that upsampling should only be used if their upsampling is better than what the DAC forces on you. They are of the mind that upsampling for any other reason adds nothing. I would not start this discussion over there unless you want to get lectured.
The idea of upsampling from 44 to 176 rather than 192 is that 176 is an integral multiple of 44 and 2 out of 4 of the original data points can be used. When you convert from 44 to 192 then then all of the data points will be calculated ones. Of course, the result also depends on how you DAC handles 176 versus 192.Some DACs have separate clocks for the two sample rates (44 versus 48) and some have just one clock and generate the other rate mathematically.
The answer here has to be that implementation matters and that you should try it in your system and see if you hear a difference. The results are dependent on the software converter used and the DAC. That said, unless there is really flawed component, the differences will be more subtle than dramatic.
I will say, that there are people that will adamantly state that there can be no difference between 16/44 and the upsampled version. However, I am afraid these people live in a theoretical world, not a practical world where DACs are not perfect and brick wall filters do exist.
As to DSD, there is endless discussion about whether DSD sounds better than PCM. Once again, in my mind, at this point, it comes down to implementation of the DAC rather than whether one format is intrinsically better than the other.
Sorry for being so verbose. This is a topic that I have spent a lot of time thinking about. And, believe me, there is no consensus on it. It trite, but let your ears be your guide.
