Most every modern digital communications protocol uses the AMBE or IMBE vocoders (voice encoders) that are meant to compress the audio into very low bitrates which communications systems can tolerate (because realistically even today's best hardware is still rather "low fi" in terms of quality and the transmissions are still analog in nature). If you're getting a good signal meaning DSD+ shows at least 50% on the title bar for the decoding level, and you're not seeing a lot of errors during the decode (use the -v4 switch and it'll show you more info aka be more verbose in details, errors will show an e: and various letters/numbers after it, the more you see the more errors there are) then what you end up hearing produced by DSD+ is about as good as it'll get without having a dedicated scanner for the decoding (and that's just P25 Phase I/X2-TDMA and now Phase II as well).
Basically that means that all digital protocols are going to have some level of "robotic" quality to them, regardless of what vocoder or system they're using. I've heard some really nice P25 decodes in my years, and the newest Uniden scanners do a really great job, but I instantly know they're digital decodes and they just lack that quality of pure analog voice quality that has been the mainstay of radio communications for so long now. The vocoders are getting better, however, but they still have a long long way to go in my opinion. I hear LAPD stuff these days (streaming from Broadcastify) and it just sounds like utter crap to me, one of the worst sounding systems I've ever heard, but that's just me.
Volume can have a lot to do with the decode as well which means you'd have to adjust the appropriate Windows mixer to get the signal being fed to DSD+ at a decent level (say at least 50% there as well). Mind you that the volume level and the decode level are not one and the same thing; they
are related but not on a point to point basis meaning if you raise the volume by 5 (from 45 to 50, etc) that doesn't mean the decoding level will rise by 5 as well - it could jump 20-30 percentage points higher, actually. It's all about getting the right levels, seriously.
And of course with DSD+ there's the option to tune the decode based on a given system and then find particular switches that can allow for far better and more accurate decodes of that given system, and once you get those parameters you can then create a simple batch file (just a single line or two of text, really) that will start DSD+ for you and configure it with those parameters without having to type all that in each time beforehand. Use dsdtune for auto-tuning to get those parameters, you can find it here:
https://github.com/dreinhold/dsdtune
Good luck...
