Tony,

I'm just an idiot (a happy idiot, but an idiot nonetheless) so I don't know how such nebulous processes as quantization and decimation are specifically capable of improving the noise floor (and consequently the MDS and dynamic range) of an ADC, but they are.

I found a thread in the Flex Community forum that speaks obliquely to that end - the Flex engineers stated that they were able to improve the Flex 6300's IMD DR3 figures by anywhere between 7 dB and 16 dB (depending on band and preamp settings) in the latest SmartSDR V2 software release.  They didn't reveal how they were able to do that, declaring that to be a company secret, seeing as how there's a bunch of other folks (and companies) who'd also like to know.  :)

Here's that thread:

https://community.flexradio.com/flexradio/topics/v2-improved-dynamic-range-for-flex-6300?topic-reply-list%5Bsettings%5D%5Bfilter_by%5D=all

73,
Mark  KF5VQY

On Fri, 29 Sep 2017 14:39:06 +0000
Tony_AD0VC wrote:

Maybe somebody else will comment on the assertion that ADC width
doesn't matter.

Beyond a certain point it doesn't, but you need enough bits to hit the
dynamic range needed, roughly 6dB/bit.

The upper limit is set by the reference voltage the ADC uses, the lower
limit by internal noise and by the size of the LSB. If a 14 bit ADC is
able to give a receiver noise floor that is low enough then a 16 bit
ADC will not make it any lower because other sources of noise become
dominant.

--

Brian Morrison

Noise floor isn't the only consideration.  Dynamic range is valuable when a weak signal is adjacent to a very strong signal, or when an extremely strong signal would otherwise simply overload a receiver front end.
--
Mark  KF5VQY