You said:
“Dan.
OK, you know what the problems are, would the ideal chip have more bits, if so, how many? How would you get rid of some of the problems, besides making it easier for a design artist such as yourself to selectively hang stuff off the chip?
Do we really need or desire a full 24 bits? What would happen if you had that? Would you still be in business, would your life be a little easier from the design aspect?”
I say:
I do not view what I do as “selectively hanging stuff off the chips”. I believe I am the only one that designed manufacture and sell some AD and a DA made from scratch – resistors based design, with a lot of new concepts. Some of the concepts have been copied by competitors and some will be later. So I still think you are weighting things a bit heavy on the IC side, and light on the equipment maker side, at least for quality state of the art gear.
But yes, for the main market, the IC makers are the key, especially when it comes to consumer gear.
Do we really need 24 bits “end result”? I do not think we need 144dB dynamic range. How silent things get? Say a sound proof room. How loud? Say 10 feet behind a 747 jet engine? It is still not 144dB…
But one may want to have a little margin over what you need in the real world. What do we get today? Take a state of the art mic preamp, and you find that for the most part we are around 125dBu (plus minus a couple of dB) noise floor (equivalent noise floor, reflected to the input after accounting for some gain).
The question is: how much gain do you use? Say your setting is 30dB gain, and that effectively gets your signal peaks to full scale of 24dBu. So we have 125+24-30= 119dB dynamic range, or about 20 bits. Say you have a hot mic and the singer is yelling at it so you only need 20dB gain. We now have 125+24-20=129dB dynamic range or about 21 bits…
So basically that is how you determine what is needed. You may wish to adjust my figures a bit, but we are not anywhere near 24 bits in our capabilities. The improvements will have to be done at the bottleneck, which is both the microphones and the input stage of mic preamp. We are limited by analog noise sources. A 62 Ohms at room temperature generates 1nV/sqrtHz which ends up as .14uV of noise (0-20KHz bandwidth). A 1KOhm yields about 4 times as much (over 1/2uV noise). So a mic that “looks like” a 1KOhm will need to put 1/2V signal “to be” 20 bits. For 24 bits it would need to put out 8V! Of course the other analog noise comes from semiconductors, and there is little out there that goes much less than 1nV/sqrtHz. There are some costly “tricks” to improve semiconductor noise by some, but than the resistor noise is still there…
I can think of one application where you could do away with such (microphone) constrains – synth sound source, but I just do not see it as a good reason for 24 bits (no vocals and really limited sounds).
So the limitations are in the mic and the preamp. Those are the bottleneck today, and they are analog. There are physical reasons why that semiconductor noise figures have not improved much in 25 years (I mentioned that 1nV/sqrtHz noise voltage or so, and there is also a very limiting noise current in the few pA/sqrtHz).
In the context of moving towards 24 bits:
You asked about what the IC makers can do? It is the ANALOG low noise area that needs improvements – it will advance both the mics and the preamps. Compare to that very difficult undertaking, adding bits on a compute engine is “a walk in the park”.
In the context of exsisting mics and preamps:
The AD and DA IC’s are not yet 24 bits, though some are specified as “20 bits A weighting”. They could and will improve it, and the job will be easier when we get rid of that 192KHz un needed burden.
In the context of what we need:
It would be great to have 21 bits AD noise floor, non weighted, out of an IC. Do we need more? I am not convinced we do. The final product is often less than 100dB dynamic range, and that can sound great. So that extra 20-30dB are there because it gives you margins to overcome all kinds of deterioration due to processing (gains and boosts and mixing and so on). This is not my area of expertise. I am an equipment maker.
You ask tough questions, thus the long answers…
BR
Dan Lavry