Ted,
For making a DAC, what would you do (differently) if you had:
Bonus question: What is the most damaging component in a DAC? Said another way, what component in a DAC just can’t be made better and would be if it could be improved?
Peace
Bruce in Philly
I’d need an infinite amount of time to consider the options. 
The most obvious thing that comes to mind is generating all possible bit streams and choosing the one that gives the output that’s most faithful to the input. I’ve coded a version of this with look ahead to allow reasonable pruning of the massive search space and it does work, but there are always bit streams whose bad features aren’t apparent until after the look ahead window.
Something more practical is doing a two level DAC where the deviation of the analog output of the first to it’s input is used to guide the second in doing a better job.
There’s always environmental noise as well as self generated noise and both of these are next to impossible to completely ameliorate.
Also, of course, “perfect” upsampling filters.
But would you at all be able to capture all differences? Should be with a digital signal, but why then did people in the 80s think they already got the identical result?
Is the only problem that we don’t get the identical signal yet, or is it also that we don’t know all parameters yet to check for?
E.g. if we look at water coming out of a tube and check for identical water flow, but forget to check for radioactivity because we don‘t know yet what it is.
If the analog is bandlimited then you can get arbitrarily close with digital of an appropriate sample rate.
At a more pragmatic level the goal of a DAC is to reproduce the analog that when digitized matches the DAC’s input. The difference between the input and output of the first DAC can be done either in the analog domain or the digital domain. In either case it’s a smaller signal (less information…) than the input and doesn’t need the same accuracy/care that the rest of the DAC needs. After all if it could be perfect then the DAC could have been perfect in the first place.
The point of using one DAC to help correct the other is simply a good way of partially correcting any systematic errors in the DAC. But it’s costly and could do damage to the signal if not implemented well enough or if the system is too noisy overall.
Yes I see, if we have unlimited capacity we can simply do endless runs of less exact comparisons.
And I understand that the primary goal of a DAC seen isolated is to equal the input signal.
This definitely will sound best if the input signal would be perfect (which it is not). I guess that’s the only reason why for some a tube buffered DAC makes sense, although this is just done to cosmetically add something to the sound which is perceived as advantageous by them due to the unavoidable imperfection happening before and after a theoretically perfect DAC. The.more perfect the surrounding before and after a DAC get, the less good sounds a compensating DAC design…just as it is with any compensating gear being dropped as son as the surrounding strongly improves.
So far just a theory.
Sounds like a quantum dac is needed. Someone go ahead and invent one.
The interesting thing for me to think about regarding such topics (and this has nothing to do with digital in particular) is always, if our assumptions to form a goal are the right ones.
I think they are. Because it makes sense to stay accurate at least until deciding to add or filter something (either by “voicing” or other measures)
But anyway it often seems crazy what effort we put in playing back a signal, which seems much more compromised in the process prior to having a ready produced media in our hands, than in top high end playback chains thereafter. And how we try to replicate the incoming signal (which is so compromised by restrictions of mic’s etc. etc.) as exactly as possible.
Imagine if someone had tried to build the best amp ever to amplify a grammophone and the best speaker to play the music. Would the one, most accurate to the input signal, have provided the best sound? Rather no.
I somehow think we’re still in a similar position, but on the other hand we’re probably far enough to take this input signal as a good enough reference. But for many, I’m sure e.g. tubes are used to compensate and add what they still miss from this accuracy.
I am not quite sure where you guys are going with this, but honestly, after a bajillion years in this hobby, yes I am that old… I have come around to saying YES to enhancement. No, I don’t need a blue pill. But when you understand weaknesses, or damage done in the chain, go ahead and repair it. While Dr Heisenberg will tell you it is impossible, improvement is still possible.
I am OK with it. Go ahead and extrapolate those missing points… filter out the baddies… heck throw a tube in there to add some loving distortion. I just want beautiful music.
Peace
Bruce in Philly
When I started this project I wasn’t looking forward to “voicing”. But, thank goodness, almost everyone liked things more when they had more fidelity to the input. That means I can simply concentrate on objective criteria for judging possible improvements. If people want other forms of coloring, treatments, eq, etc. they can do it before or after the DAC.
I always find such comments very reassuring. 
Amen, brother Ted.
Puts me in mind of Quad’s “current dumping” amp design though doubtless many others have implemented similar, in analogue and (maybe) digital
That’s fine!
We just must admit, that most every cable, streamer, amp and speaker manufacturer sees it similarly. So if we take all of them serious, there’s no option to voice anyhow, as all the equipment is accurate to the input. We know, true life tells a different story and each of them is colored and sounds different 