New Zoom recorder. It’s tiny.
3 Likes
I am looking for a simple recording solution for my master out. Has anyone here experience with the Zoom F2? Preamp and recording quality? (I have also been looking at more fully featured multitrack recorders such as the Zoom F6 or the Sound Devices Mixpre2 series, but I simply don’t need more than one stereo input.)
Edit: just found out that it records only in mono… 
Get a h5. it has stereo in as well as the stereo mic, but you can later buy an expansion to have 4 inputs so two stereo pairs if you want. H4n is good too, better preamps I think but it’s not modular like the h5.
I have a h6, upgraded from h4n but I have noticed that the h6 is a bit more noisy. Not a big deal when recording synths, but sometimes audible when doing field recordings.
1 Like
What’s going on with this 32 bit thing? Very sceptical about the lack of gain control and their claims that you can boost the signal indefinitely with no noise
32 bit floating point gives you a little more than 1500 dB of headroom. Apparently the power output of the entire observable universe is around 500 dB.
I can’t imagine the ADC will come anywhere near the total range of 32 bit float, but you will probably get more range than the typical 24 bit recorder.
You will need to normalize in post production to bring the levels of what you’ve recorded up. Think of it like HDR, but for your ears.
4 Likes
Yeah the key to this device is it records in floating point.
To get an idea of how floating point makes a difference, if you never took a computer science class, think of the difference between regular numbers — 0, 15, 26034 — and scientific notation — 1.71E0, 3.62E-34, 4.75E93 — notice how the scientific notation numbers can go from very small to very large.
Here is a very good article that lists all the floating point field recorders available today:
The F2 is included in this list. It also talks about what floating point is too.
They also give a link to a place that talks about software that groks floating point:
For field recording the huge dynamic range lets you deal with setting volume for the shouts and whispers later in production. Delaying decision is good. Give me floating point for field recording.
8 Likes
This problem has bothered me for a long while, ever since reading about the Sound Devices floating point recorder. I’ve worked out in my head how i’d design an FPR to get around this problem, and improve things.
Pretty sure there is no such thing as an “analog to floating point” converter chip. So they are using the customary good quality “analog to integer” converters, and then converting that data to floating point.
My trick to solve the dynamic ranging problem is to run more than one “analog to integer” converter device in parallel on the same analog signal.
One approach is to put in front of these converters switchable audio gain circuits. When the device sees the range of the ADC it is using move up above a threshold, it switches to a backup ADC with a gain adjustment set to cover that higher range, and changes the calculation to convert integer data to floating point so the measurement stays continuous. The same can happen when the audio signal volume goes down.
With two of these devices you could “ping-pong” between them and move up and down a long range.
You could also have multiple parallel ADCs all with different audio gains, and moderate between them to avoid the analog clipping. This second variation on the idea avoids switchable gain circuitry. A lot of existing ADC chips already come with multiple converters, so you just put different audio gains in front of these separate ADCs.
This does not get around the dynamic range problems of your input transducer. You obviously want very high quality microphones.
If i wanted to even get around the transducer dynamic range issue and had big money like a movie studio, i’d have multiple microphones, each with different volume characteristics, and switch between them. Or even better record them all synchronized in digital and pick the sound gains later. There is then phasing problems even with the multiple microphones tightly placed, so add some DSP software to deal with that.
Ending this post here.
1 Like
Looks like you could connect an external stereo input if you had something nice. Pretty compelling combo with some stereo in-ears.
Well said. At the price point of the Zoom, it’s an easy impulse buy for my next vacation. (Remember those?) will be fun to play with and see how the design decision tradeoffs land.
1 Like
If these new-ish recorders use 24 bit chips, and then convert the files to 32bfp… that’s what already happens when you use Zoom’s 10 year old recorders and then load your audio files into free software like Audacity. You use a 24 bit integer ADC chosen by Zoom, and then when you open your audio file on your computer, it gets converted to 32bfp. Same difference.
Scientific notation is only more efficient if the number ends in many zeros (or starts with a decimal point and many zeros), which you can then truncate and represent with an exponent. Recording in 24 bit, and then converting to 32bfp, all you can add, in good conscience, are zeros, yes? Or else they could convert to 32bfp, then run it through their fx dsp, and get meaningful bits that way, but that’s not the same as actual 32bfp recording. It would be disingenuous to advertise this as 32bfp recording. Hopefully, that’s not what’s going on.
No it’s not.
What’s the difference, then? (between what the new Zooms or other 32bfp recorders do, vs recording with a 24 bit integer Zoom recorder and converting to 32bfp in Audacity)
I’ve got a couple 24 bit Zoom recorders, and I convert the audio to 32bfp when I process it, so I’m wondering what would be the advantage of a 32bfp recorder, if it actually records at 24 bits.
Maybe these are 2 different questions?
Apparently these new 32 bit recorders have a huge amount of headroom and are impossible to clip.
2 Likes
Not if they use two 24-bit ADCs per channel, with different gain, and then choose which 24-bit resolution signal to convert to a 32-bit float.
As described above:
When the input is quiet, you use the high gain ADC’s output and have a small exponent for the float. When the input is loud, use the low gain ADC and a large exponent. They gives you 24 bits of resolution, but a higher dynamic range because you can scale the window that you capture in those 24 bits.
1 Like
Think about what you do manually to get greater dynamic range in two different recording environments. Sometimes you turn the gain way up, and sometimes you turn it way down. The FPRs do this on their own and very quickly and in order to store that greater range they encode the data in fp.
2 Likes
Ok, so it’s 2 ADCs, 1 with a pad first, and then it comps from whichever is most appropriate automatically, instead of you doing that manually, afterwards. I didn’t think Sound Devices was originally calling that “32 bit floating point”, but just advertising it as a useful safety feature for very dynamic field/location recording.
Not a big feature for my personal use then. Not enough for me to upgrade, at least. In light of the explanations, I still think it’s disingenuous to call it 32bfp recording. At least I have a little better idea of what they mean, now.
I don’t. A 32-bit float has 24 bits of significand (the integer part) and 8 bits of exponent (how big or small the integer part is scaled up or down). So if you captured more than 24 bits of input resolution you’d need more than 32 bits to store that losslessly as a floating point number.
Edit: the 24 & 8 figures assumes they are using IEEE 754 single-precision floats. Given that 1500dB is impossible to achieve, it might make sense to use a different format with 28 bits for the significand and 4 bits of exponent, or some other split. That would allow capturing 28 bit resolution, and still a high dynamic range. But I would assume (from a position of zero expertise) that commodity ADCs and processors probably only support 24 bit capturing, and IEEE 754 float format.
2 Likes
I guess I could get similar benefit by running a Rube Goldberg contraption of the Zoom R24 and L12 in parallel, after a splitter, but if I really needed to do that, I’d probably just get one of the 32bfp recorders instead. 
Just give it a few days to consider, i think you are close to getting the difference. It’s real.
2 Likes
Thank you! That’s the part I was missing. I knew that, too…
Still, since it’s using either/or and not both at the same time, idk about calling it that. Technically, I guess they can. I wasn’t planning to try to stop them, either way. But idk…
1 Like