Why Manufactures Use Multi-Bit Delta-Sigma Modulators?

I was reading this article and SNR for a 5th order delta-sigma modulator with x64 oversampling is 149db/24.5bit (p.7). That is more than enough. I’m using Neutron Music Player which has a 5th order DSD converter that should have the same SNR.

So why bother with multi-bit DS modulator and dynamic element matching when 1bit is, or seems to be, ideal?

I know it’s not just the quantization noise that determines SNR (in fact, the article says that, too), but still.

I appreciate people that help me understand it, especially @MazeFrame and @Polygonhell who helped me before.

Honestly I don’t have a clear answer to this.
The short version SNR is not a sufficient metric to understand how a piece of equipment sounds.

The problem is they are looking at steady state signals.
Pretty much any DAC sold today (or 20 years ago) can reproduce a 1KHz Sine Wave at 2V with > 100+dB SNR, so why pick any solution over another?
Audio companies do actually listen and blind test their solutions…

The reason DS DAC’s became prevalent is they were an obvious match for DSD. DSD while remembered for SACD really came out of a need for fast, high bit depth A2D conversion in studios.
It was “difficult” to make fast high bit depth A2D converters in the 90’s, DS A2D conversion “solved” that problem, and SACD spun out of that.
High accuracy R2R DAC’s even on a chip are difficult and as a result expensive to make, so over time the DS chips killed them off.
Now they only really exist in the commercial space for applications like medical imaging - which is where Schiit gets theirs from.

As I understand it, oversampling introduces time domain artifacts that are not present in standard frequency sweep type tests. These are most visible in impulse response measurements, but also visible in Square waves as ringing artifacts.
I SUSPECT that by directly decoding the top few bits of the signal, you achieve 2 things, first you reduce the magnitude of the ringing artifacts, and secondly if you retain the oversampling frequency, you shift them further up in the frequency range making them less likely to be audible, and more likely to be eaten by the analog output filter.
What effects these things have on how it sounds I couldn’t guess.

But DAC’s aren’t just the DtoA chip they have analog components in them, when people talk about implementation, it’s the circuit the D2A is in and how it’s utilized.

When you look at higher end DAC’s the choice of converter technology seems to be less of a driving factor, there are a lot of R2R solutions, but fore example the Linn Klimax DAC which is very expensive and highly regarded uses off the shelf ESS converters, The Lampizator DAC’s all recently swapped from R2R to DS solutions.
FPGA DAC’s like the chord are as I understand it are pure 1 bit solutions with MASSIVE oversampling rates and “clever” digital filters.

My guess given the variety of solutions at the cost no object price points, is that it’s about trade offs.