first off, thanks, Paul for chiming in on the mixpre - good info as always
back to the zoom:
I decided to do some tests on the zoom, testing recording at 24bit vs 32bit FP at various sound levels, feeding a signal of a properly recorded audience recording from my DAC (Mytek brooklyn, claimed dynamic range of 130 dB)>balanced out>zoom set to mic-in, and adjusted level in 20 dB increments
I knew at extremely low levels i would run into the noise floor of the DAC, so i tried 2 different methods:
1. playing a signal from the computer to the DAC at constant level, and attenuating the DAC output via the stepped pot volume knob (the mytek reads exact dB of attenuation). For this test #1, I used Scott Schneider's most excellent 4015>portico pull from Dick's 9/1/19. it was a 16-bit version (which shouldnt make much difference as any noise would attenuate relative to peak signal)
2. making an file in soundforge that repeated the same 2 minute clip, which i normalized to 0dB (peak), -20 dB, -40 dB, -60 dB, and -80 dB. I then played this back with the mytek set to constant output level. The source material used in this case was an actual 96K 32bit float file i recorded at a concert. After doing the above normalizations, i took the -80dB segment and normalized it back to 0dB. As expected, the 32-bit float container handled the data appropriately and no noise was introduced. That doesnt mean that the -80 dB signal fed to the DAC and then to the zoom is free of noise, but whatever the DAC is putting out should hopefully be recorded by the zoom giving us some vision of the input noise floor and whether recording bitrate affects it. In fact i had to set the volume of the Mytek to -19dB to not overload the zoom in 24 bit mode, so i may have compromised the dynamic range there a bit,so i did test #3...
3. same as #2 above but with mytek 'wide open' at 0dB, i skipped the sample normalized to 0dB, and did -20,-40, -60, -80, and -100 dB samples (the latter of which should roughly correspond to the same ~ -80dB on the zoom as the lowest level sample of test #2)
another variable i encountered was, in mic-in setting, the minimum level the trim in 24-bit mode can be set to is +12dB. so the levels on the iso are 12dB higher. This difference seems to make no difference on signal/noise however
4. same as #2 and #3, except zoom was set to 'line in', and trim was set to -8dB (the minimum) in 24-bit mode. samples normalized to 0, -20,-40, -60, -80, and -100 dB were played back through the mytek set to 0dB volume. The Zoom was input was overloaded a few dB in this case on the 0dB sample. 'Exceeding input volume' was displayed when recording in 32-bit float
5. same as #4, except substituted Tascam DR100 mkiii for the zoom
6. same as #4 except substituted AD2K for the zoom
7. same as #4 except substituted V3 and AD2K for the zoom, and recorded the -60 dB sample with 60 dB of gain on the v3, and then the -80 dB sample with 70 dB of gain on the v3
----------------------------------------------
i normalized volume of all samples with peaks close to zero, then saved as mp3s.
for the case 1 files i went straight across and normalized all of the samples (0, -20, -40, -60, -80)
for case 4, AD2K, and V3 AD2K, i just did -60dB and -80 dB. both the -60 and -80dB samples have a few seconds of (what was) digital silence on the original recording, after each sample
comparing the V3 files to the ones with no gain, it was apparent i was running into the noise floor of the DAC.
That said, all the files sound darn near *identical* to me, and the noise floor in between samples is basically identical on both the 24 and 32 bit samples. the -60 dB samples are passable in both cases, the -80 dB samples are excessively noisy
on hot signals the zoom overloaded at the exact same points (even though with the trim, the 24 bit files were hotter. the 32 bit files had lower level but the same harsh clipping
Im not sure how valid the test was, due to the inherent noise floor from the DA. I think i need to do another test with the lowest sens mic i can manage. perhaps a DPA 4006A or 4007A with the 20 dB pad on. Which do you think would be more useful, the 9mV/Pa 4007 with the 24 dBA noise floor, or the 40 mV/Pa 4006 with a 15dBA noise floor. Both those specs are before the 20 dB pad.
what other gear would allow me to run a clean-ish low level signal in? maybe come off the mytek with a hotter signal and use some XLR inline pads?
in any case, it seems that for all *practical* purposes (i.e. guessing recording level within 40, or heck even as much as 60 dB), the 32 bit doesnt offer additional resolution
files for your listening displeasure. Even though they are mp3s it might be best to open them up in a wave editor so you can get a visual for the samples at which you are looking
https://drive.google.com/open?id=1taG_IOlFEnlXZsD0bmveKt6XWLYmefAa