There are a few ways of doing a DSP speaker, but obviously there isn't a consensus 'great one', or else we'd see everyone using it. Something like JRIver makes a good platform, but struggles handling external sources. The Hypex DSP amps are attractive, but if you want to feed them digitally you need a digital preamp/switcher which doesn't appear to exist at the moment (except maybe for Meridian - I'd love to have a slightly updated 518) and they're IIR only. The MiniDSP units are popular as they're cheap, but are only uncomplicated if you're ok with analog input - as soon as you want digital inputs you're back to the switching problem; plus, they all SRC to 48/96 which isn't the worst thing in the world but (see below) I'm not convinced it's ideal. The MiniDSP SHD or SHD studio is probably pretty close to ideal if you can live with a 2-way system and accept being limited to IIR filters + Dirac, but putting all that effort into something and being limited to 2-way is a bit disappointing.
I don't have any experience with the MiniDSP OpenDSP units, but I do have a NanoDigi that I'm evaluating for a DSP speaker project. My current gut feel based on some A/B testing is that the SRC on the input of the NanoDigi is NOT audibly transparent, but I need to do more due diligence before actually concluding that. The NanoDigi is IIR only, unlike the OpenDSP. I do wonder whether the OpenDSP unit really has enough FIR capacity to be viable though - only 9600 taps shared across all channels. Assuming they're not providing any clever multi-rate capability, it's tough to see how you could do woofer xovers with that few taps. It might be viable for a setup with IIR filters between woofer and mid and FIR between mid and tweet, though. (for reference, when I was running BruteFIR in the past I'd use 16k to 64k taps per channel)
My current project is a pair of Kef Q150's that I want to add a pair of woofers to for my desktop use, but I've been putzing around with DSP crossovers for almost 20 years with varying degrees of success. More 'lessons learned' than true successes, sadly, but this time - THIS TIME - it'll be different
The short form blurb is that my current thought/plan (assuming I don't go with the NanoDigi) is to use a Linux pc running BruteFIR as the DSP engine. I'll probably run jackd for connection management, and will have to write some relatively simple software to handle remote control for input switching and potentially volume control. (I'm a software guy, and have done some jackd clients etc in the past so in theory this is only a matter of finding the time)
For output, I'm currently think of using the MiniDSP UDIO-8 running into a setup of the Khadas Tone Board plus JDS Atom for each output. I'm only looking at a 2-way system, which means this will run $300 for the UDIO-8 plus $200/way for ~700 total. Not too bad. If you're looking at a 3-way or 4-way system it might make more sense to look at the Okto Research Dac8 Pro (which isn't shipping yet, unfortunately). It's only ~1200 and appears to outperform everything out there AND provides global remote-controlled volume control. (the inclusion of the Atom in my setup is simply to provide a way to gain-match each channel fairly closely so that digital attenuation can be kept moderate to avoid too much compromise of performance. In my past DSP setups, the woofers have generally been a problem, requiring a fair degree of attenuation of the mid/high)
The key factor in both of the above setups is that the *output* runs at native sampling rate, but there is SRC on the digital inputs. This suits my specific application where all of my ripped content is 44.1 and that is what I'm prioritizing, so I can play my 44.1 content via MPD or Squeezelite etc directly into jackd with no conversion, but still handle external digital sources transparently due to the SRC on the inputs - all that is needed is to change the port routing in jackd. Without SRC, you are faced with having multiple filters sets - one for each sample rate - and having to re-initialize the filter engine any time you change rate.
All of this 'should' result in something close to a digital preamp w/ crossover functionality - similar to the MiniDSP SHD, but much more powerful (aside from lacking Dirac) but unfortunately less user-friendly - generating the FIR filters is always going to be finicky, and getting ergonomics / usability is always difficult. My ultimate goal is something stable/useable enough that I could conceivably consider it for a living room system (I have a vision of a DIY project emulating the look of the Sonus Faber Stradivari which I could only take on as a DSP project)
I'm a big fan of the Dayton subs. The RSS210HF worked quite well IMHO when paired with the NHTs, and I'm hoping to try using them with my current Kef Q150s. For a bigger room, the RSS265HO-4 does amazingly well in only 1-2 cu. ft ported. For a music-only system, I think these are very good choices as long as you can absorb the box size - they'll be bigger than the Xdw I would imagine. I also have 2x RSS390HF in our theater space, and they're great as well but massive overkill for an audio-only system.
BTW - I'm not sure whether you've seen this. Audioholics already measured the transfer functions of the digital filters in the Xds system, as well as the acoustic output of each driver individually. Not the best image, but I'm guessing you could use it to get pretty close using RePhase to generate the filters (I'm almost certain the Xds filters were linear phase)
https://www.audioholics.com/soundba...xd-loudspeaker-system-review/XdAOut.jpg/image
Main page with measurements is here
https://www.audioholics.com/soundba...review/nhs-xds-measurements-and-analysis.html