The way a multi-bit DAC works is that you have a bunch of voltage or current sources - one voltage or current output for each bit.
Each bit is twice the value of the bit below it (so 1v, 1/2v, 1/4v, 1/8v etc... or 100 mA, 50 mA, 25 mA, 12.5 mA etc...)
The "ladder" is simply the way they've chosen to generate these voltages.
Connected to these in some fashion are a series of electronic switches.
The switches turn on and off to "use" each bit by connecting it to the output - or not.
(These switches are designed to be either ON or OFF... so you can't really fiddle with their values much beyond that.)
It is these voltages or currents that need to be very precise.
If you know a little bit about voltage dividers, you'll realize that "doing this the obvious way" would require a whole bunch of resistors of all different values.
The "ladder" part of the description refers to the fact that someone a long time ago figured out a way to derive these voltages using a whole bunch of resistors of the same value.
This is a huge deal because it's easy to buy precision resistors in big batches of the same value - and to get them really perfectly matched... so it makes getting all those different values correct much easier.
(A typical "ladder network" uses all resistors of two different values.... you can Google for details of exactly how they're connected.)
There are also similar multi-bit DACs that use arrangements of capacitors rather than resistors.
However they ALL share the fact that, at some point, there are switches that must change state as a new number is read in...
And ALL of them have the problem to some degree that the switches cannot switch exactly at the same time...
And, because of the disparity, there is always at least a tiny "rough spot" when the switch actually occurs.
There are various ways of keeping this from affecting the output too badly - and many purpose-designed audio DAC chips have these built in.
And, if you use a chip that does not, or if you build your own DAC using separate parts, it's one of the design problems you have to solve.
There's not much you can do to make the resistors more stable or accurate... other than make the out of the right materials... and control their environment.
When resistors are made, they are often LASER trimmed (which, just like it sounds, means that a machine sits there with a meter and burns little bits off the edge until they measure precisely what they should).
Most resistors vary with temperature, you can shut your resistor in a little box with a thermostat (that's what a "crystal oven" is).
Also, by carefully mounting all of the resistors in your ladder near each other, or on the same plate, you can try to make sure they stay at the same temperature.
(Since all of the resistors in a ladder are the same, they are probably affected the same by temperature; so, if they all stay at the SAME temperature, so their values vary together, the value of the voltages between them will stay the same.)
There is also a reference voltage that feeds the ladder.... and you can design the circuit that makes that very carefully and make sure it's stable.
(But, since everything uses the same reference voltage, if that drifts a little over time, the output level will shift a tiny bit, but it won't cause any real harm.)
Virtually all modern DAC chips these days are actually multi-but Delta-Sigma implementations....
Which means that they are a sort of combination....
They have one or more three to five bit multi-bit DAC circuits, connected to a Delta-Sigma section.... (there are technical reasons why they do it this way).
The way the insides of a R2R DAC works are actually pretty simple... you could draw one on a napkin with a thick pen.
Delta-Sigma chips are very complex... and each manufacturer does it somewhat differently.
Both the Sabre and AKM chips are Delta-Sigma designs, and both include a lot of extra features as well..... so they are VERY complex.
(They also work very well.)
I think you'll find that many manufacturers tend to sort of blur the descriptions....
What they're doing is implementing "deglitching" to eliminate those switching issues....
There's really nothing you can "do" to "keep a resistor stable" other than choose the right one to begin, keep the reference stable, and mount them such that they remain at the same temperature.
The reality is that none of this cool sounding tech stuff really matters.....
All that really matters is how closely the ANALOG output ends up resembling what it's supposed to.....
I read your referring to the ladder dac chips that are non-audio, which are mainly the schiit multibit dacs.
But These new "ladder" dacs are actual resistors, so that has to be different, right?
No where do I read about any "glitch" Problems with the new ladder dacs ( that not using chips.)..
In fact I only read about that with schiit dacs, which I read they solved.
When I looked, I only read about true ladder (resistor) dacs (not chip) concerned about implementing their own methods,
in to keep the resistors stable and accurate.
Please correct if I am wrong because I am way Unfamiliar with these things.
I also see that both AKM and LeSabre have recently come out with new dac chips,
So it will be interesting to see how new dacs perform with those new chips...