Oh! I misspoke. Actually what you show is just what I was thinking...well maybe. If you start with one-bit at the A/D then the noise is going to stay the same even if you "PCM" it.
- Can you do the same comparison, but up to super high frequencies? (I'm kind of thinking the PCM will decimate off the highest frequency noise but my thinking on that is blurry today-I need to use MQA to "de-blur" my brain ha ha)
Since highest PCM rate I can get out of ADI-2 Pro is 768k, it is not possible to compare to higher bandwidth. But I used matching rate families for this and not the 48k-base DSD although it works fine with ADI-2 Pro as well.
From noise perspective it doesn't make practical difference if it's like 1, 2, 4, 6-bit ADC or DAC. Modulators are blasting through full value range all the time. Sometimes multi-bit modulators are used in attempt to get away with simpler modulator implementation, but then you can usually see side-effects of it (can be observed especially with DAC chips).
If you convert either of above to 96/24 PCM, you see just a flat noise floor. If you convert to 192/24 you can see a little bit of lift in the noise floor above 50 kHz.
What I should have said is are there any non-hybrid pure multibit A/D converters any more? And if we measure those, how do they compare to 1-bit A/D?
Not on the market, have not been for many years. Last such was the said Pacific Microsonics Model Two. Which was used a lot for HDCD productions, technology from times before MQA, but technically similar. And it's maximum sampling rate was 176.4k. You can find some old recordings made with it from Reference Recordings for example, their newer ones are with DSD converters and such though. At least some of the 176.4/24 hires material you can find may be recorded with it.
Challenge for those are the similar you have for R2R ladder DACs. It is hard to get good precision and low level linearity. And you cannot use high oversampling rates because then you lose bits from the wordlength due to settling time issues, since converter needs to settle to +-½ LSB in fraction of sample period. This means you need more aggressive analog anti-alias filters which in turn means that especially your phase response gets bent.