By using a newer generation FPGA there would be more room for DSP, I’ve got a few ideas about things that wouldn’t fit in the DS/DS Jr. That’s where much of the improvement in the DS and DS Jr is from so I want to make sure that it can continue.
I’d continue trying to make cable quality irrelevant, in particular I’d try to directly address groundloop issues and lower the noise (especially from the digital part) that goes out (or in) on the cables (including making sure that less noise goes out (or in) the AC power cord.)
I’d use top of the line Jensen Transformers. I’d lower the analog output noise by doubling (again) the number of digital switches. I’d use parts with better specs where it makes sense.
I’d probably put the digital and analog into separate boxes connected by an optical link. I’d try to make sure that there was no digital hardware of any kind in the analog box (e.g. no processor (which means no display for the analog box), tho receiving a digital input signal means there’d be some kind of flip flops, etc. in there, but they don’t have to be CMOS or other noisy circuitry. (This also rules out any DSP based clock multipliers, jitter reducers, etc. that a lot of people are bragging about these days. but I never liked those anyway.)
I’d use more power supply filtering with many units in parallel to lessen the noise generated (including multiple AC transformers for the different functions.) I’d get power while staying away from the peaks of the input AC (which is where all of the typical noise generated by linear power supplies in most audio equipment is.) This will both lessen the noise introduced into the boxes but also lessen noise going out of the box when other linear equipment is most sensitive to noise.
Tho everyone wants to hook everything up to the net these days for updates, diagnostics, etc. I’d make sure that the people who don’t want any radios or 100MegHz or 1GigHz… generators (Ethernet) in their units don’t have to have them for normal operation and also that there’s the option of using simple IR in any remote.
I’d still continue to avoid things that are trendy but generate noise (e.g. RF remotes, live network connections, supercaps, clock receivers, synchronizers, PLLs (the analog box has to be the clock master.) To be clear I think there could be RF remotes, network connections, etc. but for those that don’t want to use them they have to be completely disableable when not being used and there has to be equivalent function provided without using them.
There are always points of diminishing returns, so even “cost no object” devices are built to some budget. Without a budget (and other constraints) you end up with the “everything including the kitchen sink” problem and you make the design time, etc. last “forever.”
Note that a few things wouldn’t cost much more, but no matter how one slices it I can’t think of ways to avoid significant cost for most of the items above (the most glaring thing being the output transformers.) Things like lowering noise by doubling (or tripling, etc) units obviously adds cost. Achieving the things that matter to me (like everything above) while not throwing people’s money away would take some serious design work.