Sorry for the long delay in replying. The reason a digital recorder with analog mike inputs, fixed-gain preamps + a/d can't match the dynamic range of an a/d in a digital microphone is exactly what you surmised. Professional microphones exist in a range of sensitivities from 1 to about 25 mV/Pa, with a few RF condenser microphones exceeding even that (ISTR a Sennheiser with 40 mV/Pa sensitivity at 1 kHz). That's a 25 to 30 dB range, with the noise floor of the microphones varying approximately in proportion. And some microphones can put out 6, 7 Volts or more. The analog electronics of any such recorder are doomed well before any digital encoding (24-bit linear, 32-bit float or otherwise) comes into play. They MUST add extra noise nearly all the time, if they are to be overload-proof with any invariant gain.
And since not being overloadable is the chief selling point of these recorders, under at least some practical circumstances (combinations of microphone sensitivity + maximum SPL at the mike position in whatever's being recorded) the resulting recordings can't possibly be as quiet as they would be if the preamp gain were set for the specific, expected range of input voltages. No standalone analog preamp is (or can be) both quiet enough AND overload-proof enough to allow this with any fixed gain setting.
At least in theory a preamp with multiple branches at various gain levels, integrated with an a/d converter, could do it; it would beat out any combination of separate gear at any price, and the first one to offer it would be world-famous. I haven't heard of any such thing, though; have you? There sure as hell isn't anything like that built into a $400 pocket recorder at present. Some day, probably there will be chips that can do it for $10, and all existing standalone mike preamps and digital recorders with analog inputs will suddenly look as quaint and obsolete as 1880s telephones look to us today. But for now, if no input gain control is provided, then the only way for the analog circuitry to avoid overload is to set an input gain so low that it cannot possibly produce the best advantage from a noise standpoint.
If someone would like to dispute this, please measure the dynamic range of your recorder via its mike inputs through to the digital data stream. Find the overload point (as an input voltage), record a tone at the maximum level--then remove the signal generator, put an equivalent resistance across the mike inputs, and record some of that. Load the track into your favorite recording software and measure the noise floor (in the 32-bit data) as a number of dB below the maximum level. If the ratio is greater than about 144 dB (approximately the dynamic range afforded by 24 bits, ignoring crest factor), I'm wrong and you can tell me so to my face right here.
But no 32-bit recorders that I'm aware of are touting any dynamic range specifications (via their mike inputs or otherwise) that go anywhere beyond about the 16-bit level. They would surely do so if they could--it would be epoch-making news.
Topic: Schoeps CMD-42 (Read 7069 times)