thanks
@amirm for the tests
both the zoom and the tascam tested recently are my units, FWIW
I've never used the zoom as a USB interface, only as a field recorder in 24/32 bit mode to 48K/96K files. the menu is rather complex so its possible something is goofed. I dont have the unit in front of me obviously, and with the pandemic i havent used it since january so ill try to root out anything that might be an issue.
from the
manual, we see that certain USB modes are limited to 16/48.
Based on this comment
Yes, something is wrong there as this device won't even support sampling rates above 48 kHz. I am testing something else but will go back to this a bit later. Thanks for catching it.
it appears we are operating in that mode here. You can confirm this by looking at the USB mode selected in the menu, on pg 141 of the manual
first of all- what we know about the input topology:
-Amir mentioned this is a 32-bit unit masquerading as a 16-bit, but in reality, it is 32-bit
float so the best we could hope for is really 23 bits of resolution in a given gain range. just to distinguish from a 32-bit fixed signal with 192dB of significant bits
-the unit uses dual ADCs tailored for different gain ranges in an effort to increase dynamic range. Perhaps the distortion that Amir is seeing is crossover distortion between the two ADCs, or some dither/LSB noise from the 16-bit operation
-the least-sensitive ADC (or the input stage directly in front of it) is fixed to a max input level of +4dBU. As Amir has noted, in my tests clipping was severe above this level. When 'line in' is selected a 20 dB pad is applied to the front end that raised the noisefloor by about the same 20dB in my tests. switching the line-in pad also raises the input impedance slightly from 3K to 5K
-with its relatively low +4dBU max input, dynamic range is constrained between that and the unit's EIN, which is stated as -127dBU 'A-weighted', so likely a few dB higher unweighted. So no matter which mode you use you are going to be limited to a practical 20 bits at best
this was the summary of a series of discussions i had with zoom support
Zoom support said:
The dual ADC is used when recording in 32 bit float and when recording in 24 bit. The Trim allows you to set the level that will be recorded to the 24 bit file. This makes it so that if the signal clips in 24 bit mode, it is due to the file clipping from the limitations of the 24 bit integer format, not from clipping at the converter. Since linear formats cannot accommodate the wide dynamic range of the dual ADC on the F6, a software Trim is necessary to set the level of the recorded file.
I said:
in terms of what you wrote i'm not sure i understand the difference between clipping because of integer file format vs clipping the converter. wouldn't both fundamentally describe going over FSD (unless you're talking about overloading the analog front end of the converter)
Zoom support said:
Yes, we are referring to overloading the analog front end.
I said:
i've done some tests recording relatively quiet source material , both line in, and mic in, and i really can't hear a difference in the integer vs float files when normalized. it seems like the EIN is the limiting factor.
Zoom support said:
As you point out here, the theoretical implications of 32-bit float are limited by hardware. With a quiet signal, I wouldn't expect you to hear a difference in the integer vs float files.
I said:
considering that in either integer or float the max input level is +4dbu, and the EIN is fixed, i don't quite understand how the float files increase dynamic range between these two endpoints. i understand how the two ADCs can have a wider range of 200 db or more, but i still can't wrap my head around how that increases the 120ish db dynamic range of the input
Zoom support said:
The preamp and ADC are going to act the same whether in float or integer. Floating point allows the final recorded file to have an extended dynamic range. So if recording in integer, the bit depth itself will limit the dynamic range meaning that somewhere in the full dynamic range of the preamps and ADC's it will have to cut it, meaning there is a ceiling and a floor which will either clip or not record. When recording float files, you can capture the entire dynamic range possible.
I said:
one more question: since the trim is acting as an attenuator, when recording in 24-bit mode would the best performance be achieved with trim set to 0dB?
Zoom support said:
Not necessarily. Because both AD converters are working all the time, where the trim is set doesn't have as much effect on the performance as what the AD converters decide to do
I said:
...or set to its maximum? would the latter be adding additional digital gain?. What trim setting would be considered 'unity gain' in 24 bit? 0dB isnt available as a setting in mic-in, of course
Zoom support said:
The trim adjustment is really just setting where within the available dynamic range you want to capture. The usual rules of setting gain in 24-bit should be followed because you still want to capture loud enough signal that it is high quality.
I said:
im assuming in 32 bit float, (where the trim is disabled), that the input range is already optimally matched to the input of the ADC
Zoom support said:
@amirm, id recommend feeding it a +4dBU sine wave in standalone mode at 32 bit float, 96K. none of the gain settings really do anything to the sound other than peg the level within the 768-dB range of the 32-bit container it is written in.
or perhaps you can do a multitone and analyze that?
ive been using R
ightmark audio analyzer for my basic tests, recording 24/96 test .wavs on handhelds (rightmark>USB>mytek brooklyn dac+>unbalanced>recorder), with the DAC set to output levels matching the input of each recorder, as long as the wav is recorded at 24/96 with levels close to -1dBFS, Rightmark can parse it and run measurements. Perhaps your AP software can do more advanced tests on imported .wavs of known source.
I've done quite a few measurements, and i find it difficult to swallow that all of these handhelds which predate the zoom by several years are consistently out performing it. I would expect the zoom to perform closer to the Tascam DR100 you reviewed, which in my casual tests is a class above all the handhelds. Yes Zoom has a reputation for its consumer-bent, but the F-series should have slightly better specs than their H series products, in theory.
Anyway heres what Rightmark can do with its 'generate wave'/'analyze wave' feature at the bottom of its main window. Note that if you record a test .wav in 32bit float you would probably need to adjust peak level to about -1dBFS and render it to a 24-bit wav for Rightmark to analyze it