Topic: Zoom F3

[dactylus]

back to the battery life for a second. I got mine yesterday, and had a show to record last night, I used it as a second recorder luckily because the band needed the tracks for a video shoot/promo they are working on. I unboxed it, messed with it long enough to set it up, no more than an hour I don't think. Did a little test recoding for 30 minutes. Then I took it to the show, and it did not make it through the first set, which really suprised me. I had a battery swap ready for the second set. External will be ideal for this, or at worst a battery change before every set. Might be my rig? Akg CK3X>PFA's?

^
What batteries were you using?

[eman]

Perhaps the PFAs are also a power draw, being a preamp of sorts? I still like the idea of having backup AA and being able to hot swap using a small phone charger instead of a big battery pack.

[detroit lightning]

I’d be interested in seeing some tests with small external batteries to see how long you could get.

My maiden voyage was a festival where I basically set it and left it all day, so I just snagged a mid sized anker pack and it would be enough to run for days. But I’m sure there’s a sweet spot with smaller batteries too.

[voltronic]

At risk of being shot down by my elders and betters, I think this very recent article explains things very well.

https://www.wired.com/story/32-bit-float-audio-explained/

In my own head - FWIW -  I think of it like this.  Imagine you are looking at a track on a DAW, in other words, a graphical representation of the sound which looks like a kind of ribbon.  In 16 bit or 24 bit audio, that track or ribbon isn't very wide.  If your recording gain in the analog domain is wrongly set, your audio could go off the top of the ribbon, or off the bottom of it, losing data and/or causing distortion on the way.  But with 32 bit float, the track or ribbon is so wide that your audio is like a stripe running in the middle of the length of the ribbon.  The height of that audio stripe represents the difference between the lowest output of the analog part of the recording system (mic and preamp usually) and the highest possible output of that analog part.  But the height of it is very small compared with the height of the 32 bit float ribbon.  It doesn't matter where in relation to that ribbon it sits.  You'll never get the audio ribbon to get anywhere near the top of the 32 bit float ribbon, nor near the bottom.  You can set the "magnification" in the case of the F3 such that the audio sits a bit above the middle or a bit below, but it doesn't matter where.  When you process it in your DAW later, you will be feeding it into something akin to a 24 bit or 16 bit bottleneck - and you can at that point shift the stripe of audio data up a bit or down a bit on the overall 32 bit float ribbon so that it goes into the bottleneck at an optimum height.

I now await others to say "Actually, that's crap" or "that makes sense" or "I didn't understand a word of that and fell asleep reading it."  Time will tell.  Gulp.

This is a very good explanation of the practical result of recording on a Zoom F-series or Sound Devices MixPre-II recorder. One significant thing missing both from your explanation and that Wired article is that you can only get these recordings to be what they are in the first place if you use multiple ADCs controlled by a DSP system which routes signals of different levels to different ADCs depending on their level before they are dumped in a 32-bit float bucket. Zoom uses two ADCs; Sound Devices use three, and I don't how many StageTec uses for their setup which has existed since the late '90s.

If you didn't have such a setup then you still would have the problem of very low levels getting buried in the noisefloor and/or very high signals clipping the ADC, no matter how enormous of a "bucket" a 32-bit float file gives you. I think of it like having both a small trowel and a full-size backhoe on site for a material sorting job that classifies rocks over a wide range of sizes. If you only had one or the other, or just something in the middle like a Bobcat, you would miss material at both size extremes and your huge sorting table with all its different size trays would be unnecessary.

///

Of course, the end user doesn't necessarily need to know any of that to get all the benefits of these multi-ADC 32FP systems; they "just work", and do so seamlessly. I do, however, think it is helpful to know that recording in 32-bit float on its own using a single ADC does not allow you this glorious freedom from setting levels. Case in point: I have a MOTU M2 interface for my PC, and it will output 32-bit float format to any software that supports it. It does not have auto-ranging multiple ADCs though, so while the 32-bit float recordings I track in Reaper give me a little bit of wiggle room, I still need to set input gain on the M2 just as carefully as I would when recording in 24-bit fixed.

The last point to remind everyone of is one that Gutbucket and a few others regularly remind us about: Even when recording with one of these new Zoom or SD recorders, you still have to make sure you are not clipping your mic/line input in the analog domain before it hits the fancy multi-ADC. If you record very loud things and/or use very sensitive mics with hot outputs, you may need to use mic pads or select less sensitive mics. Pay attention to "max input level" on spec sheets. SD MixPre-II can take much higher mic input levels than Zoom F series, but Zoom has a LINE+PHANTOM input setting you can select which allows you to record with 48V mics at LINE level, and that raises the max input level up to be comparable to the MixPre-II.

[hipporu]

The last point to remind everyone of is one that Gutbucket and a few others regularly remind us about: Even when recording with one of these new Zoom or SD recorders, you still have to make sure you are not clipping your mic/line input in the analog domain before it hits the fancy multi-ADC. If you record very loud things and/or use very sensitive mics with hot outputs, you may need to use mic pads or select less sensitive mics. Pay attention to "max input level" on spec sheets. SD MixPre-II can take much higher mic input levels than Zoom F series, but Zoom has a LINE+PHANTOM input setting you can select which allows you to record with 48V mics at LINE level, and that raises the max input level up to be comparable to the MixPre-II.

Good explanation with magnetic tape.

MP3ii has the ability to adjust the level of the microphone channel. I roughly know what level the performers will have or what I will record. For a long time I did not manage to feel all the joy from the 32FP. But a few times 32FP saved me with unfamiliar music in an unfamiliar place when I set the pianissimo gain level too high. There is no doubt that 32FP is useful for recording unpredictable events like thunder, industrial noises or gunshots.

[voltronic]

There is no doubt that 32FP is useful for recording unpredictable events like thunder, industrial noises or gunshots.

It is also a game-changer for those of us who are both performer and recording engineer, which is more than half of everything I record.

[Chilly Brioschi]

I absolutely enjoy this discussion.

Let's consider the old feature of auto-levels, ALC, where a circuit dynamically adjusted input gain to a median and circuit-built top limit.
Are the F6 and F3 making any analog input adjustments to normalize levels into the ADC ?

Is it possible to top the analog limits on the input circuit, and what are the artifacts?
Is it "splattery", "harsh", "choppy", blank or lowered in level?
Similarly, what happens at the very low end, when sounds barely break above a mic's internal noise ?

Having dynamic range beyond all hearing is great, but when it is adjusted to within the audible range, what do we actually get from the theory and grand maths?

[Ozpeter]

I would summarise it in this way - 32 bit float doesn't overcome the limitations of the analog side of your rig (mics and preamps and mixers) but it takes nothing away.  It's kind of like having the mic plugged into your ears, you could say.  But perhaps it's best to think mainly in terms of the sheer convenience of doing away with level setting.  That said, I used to delight in recording a symphony concert, and getting to the end with the peak level of the recorder showing I'd got to -1dB of full scale - and not over.  Simply aiming for -12dB or -6dB was no way to pass the evening away...  My theory that there is (in acoustic music) only one level made it not too hard to get it right.  (Mic goes close to quiet guitar, further away from a loud orchestra, SPL at the capsule ends up much the same).  Adding that theory to the theoretical dynamic range of a 32 bit float recorder means that it all becomes pretty low stress.  If I had one of these devices I would set the magnification to show that the mics were working even with empty hall ambience (for total reassurance) and probably I wouldn't look at the display again.

[Gutbucket]

In my own head - FWIW -  I think of it like this.  Imagine you are looking at a track on a DAW, in other words, a graphical representation of the sound which looks like a kind of ribbon.  In 16 bit or 24 bit audio, that track or ribbon isn't very wide.  If your recording gain in the analog domain is wrongly set, your audio could go off the top of the ribbon, or off the bottom of it, losing data and/or causing distortion on the way.  But with 32 bit float, the track or ribbon is so wide that your audio is like a stripe running in the middle of the length of the ribbon.  The height of that audio stripe represents the difference between the lowest output of the analog part of the recording system (mic and preamp usually) and the highest possible output of that analog part.  But the height of it is very small compared with the height of the 32 bit float ribbon.  It doesn't matter where in relation to that ribbon it sits.  You'll never get the audio ribbon to get anywhere near the top of the 32 bit float ribbon, nor near the bottom.  You can set the "magnification" in the case of the F3 such that the audio sits a bit above the middle or a bit below, but it doesn't matter where.  When you process it in your DAW later, you will be feeding it into something akin to a 24 bit or 16 bit bottleneck - and you can at that point shift the stripe of audio data up a bit or down a bit on the overall 32 bit float ribbon so that it goes into the bottleneck at an optimum height.

Extending this analogy- as Ozpeter describes above, imagine the relatively narrow track ribbon width of 16 bit, a wider ribbon for 24 bit, and a much, much, much wider ribbon for 32-bit float, This represents the dynamic range limit of the file format used to contain the data.  Okay, here's the important next part- There are three other dynamic ranges that represent other "ribbon widths" that need to be considered ahead of the file format.  The first is the dynamic range of the signal we're recording.   The second is the dynamic range of the analog signal path of microphones up to the analog to digital converter (ADC).  The third is the dynamic range of the ADC.  Each represents a separate domain that needs to fit within the range of each subsequent domain that follows.

Generally..
The dynamic range of the signal being recorded is smallest.
The dynamic range of the analog signal path is somewhat larger (this determines noise floor and overload level).
The dynamic range of the ADC may or may not be larger (but ideally should be).
The dynamic range of the file format may or may not be larger still (but ideally should be).

It's all about strategies to make sure the dynamic range of each domain is larger than the one which proceeded it, ideally setup such that no adjustments are required to make sure each will fit.

32-bit float recorders incorporate two new strategies to do this:

• The first is paralleled ADC's which incrementally increases the dynamic range of the ADC.
• The second is the 32-bit float file format which dramatically increases the dynamic range of the storage format.

Let's consider the old feature of auto-levels, ALC, where a circuit dynamically adjusted input gain to a median and circuit-built top limit.
Are the F6 and F3 making any analog input adjustments to normalize levels into the ADC ?

Astute observation, and essentially what the paralleled ADC's are doing, setup to do so in an automatic way. This has pretty much been determined to be audibly transparent.

The increases in dynamic range of the auto-ranging ADC and of the file format are sufficient to shift the "dynamic range bottleneck" to the domain of the analog signal chain feeding the ADC.  The bottom of that range is the noise floor, likely determined by the self-noise of the microphones.  The top of that range is determined by the overload point, or level of acceptable distortion prior to outright clipping, of something in that analog chain.  Could be anything from the microphones to the preamps to the analog input stage of the recorder itself.

With regards to the 32-bit recorder part alone, it's maximum dynamic range limits are now determined by the limits of it's analog input stage, not the ADC or file format.

[Gutbucket]

Is it necessary to dither to 24bit? What is the benefit of doing that?

Its always best practice to dither whenever reducing bit length.  There isn't really any reason not to other than eliminating it as an extra step.

Is it necessary?  Probably not.  If the least significant bits being truncated are well below the noise-floor, the random noise of the noise-floor itself essentially acts as dither.  Sound Devices doesn't bother to dither internally when storing a 24 bit file rather than a 32-bit float file on the recorder.  They do dither if set to store a 16-bit file.

[mepaca]

Just got back from 3 days at a festival. I recorded for 8-9 hours per day using pfa and schoeps capsules at +48v phantom. A pair of eveready lithium batteries lasted all day each of the days. I went through 3 pairs in 3 days. That included some playback after the shows.

[datbrad]

I want to start off by saying I think the technology that allows 32 bit float to operate in a portable recorder is amazing.

That being said, to me the benefits for the actual recording are dubious and this point has been made numerous times in this thread.

Let's start with these facts.

1. The maximum dynamic range of the best analog signal capture/amplification equipment made generally stops at 120db.

2. Rock concerts have maximum sound pressures coming from the stage/PA between 90db to 110db on average, depending on the venue.

3. Typical concert crowd ambiance averages 40db (for between songs babble and shuffling around, not cheering/roaring applause)

4. The dynamic range potential of the 24bit PCM format is 144db.

So, if you were to record a rock concert soundcheck in an otherwise pin drop quiet empty room the dynamic range could theoretically reach 120db.

If using a 24 bit A/D with levels peaking at -12db, the analog noise floor will be down 120db below that point, or -132db, with 2 bits to spare.

When the audience comes in and fills up the seats, the effective dynamic range of the music falls to 80db, since the crowd chatter and racket becomes the new noise floor at -40db.

Since -60db to -50db is the lowest level mark on the meters of most recorders, even a recording that only occasionally lights up the -40db mark and is gain boosted in post will still retain 104db of dynamic range.

Bottom line, with the same analog signal source it's not possible for a recording in 32bit float to exceed the practical dynamic range of a 24 bit recording, even with levels barely hitting -12db on the meters.

So, what's the big draw of 32 bit float in a field recorder? I see it as primarily to benefit foley, ENG, and nature recording challenges. If I was recording a frog sitting on a rock in a stream croaking, and BOOM, a crack of thunder happens, I'd feel far more confident setting levels with 32 bit float recorder than 24 bit recorder.

For concert recording, I just don't see 32 bit float being essential or substantive to proper level setting.

With all that said, if I was in the market for a 2 channel recorder I could use right out-of-the-box, the F3 is at the top of my list, but not because it's format is 32 bit float.

[aaronji]

^ To me, in the context of recorders with two (or more) 24-bit ADCs of differing gains, I think the big advantage of 32-bit float would be if you are doing mixing (post-ADC) on the recorder (where the combination of sources might exceed 0 dBFS). I suspect that there is a gain setting (maybe no gain) that would give an identical result in either 24-bit or 32-bit float for the individual channels. Maybe I'm wrong; the SD patent was pretty heavy reading!

[Gutbucket]

Datbrad-  At the risk of going OT in the F3 thread.  What you describe is essentially how I currently run F8 in 24 bit.  I do tweak my recording levels, not because I need to in order to maximize dynamic range or avoid overload, but only to get the relative level between channels about where I want them for purposes of playback on the recorder..

I think the big advantage of 32-bit float would be if you are doing mixing (post-ADC) on the recorder (where the combination of sources might exceed 0 dBFS).

This is a good point.  The only time I (rarely) have a problem with what I've recorded to F8 in 24bit mode is playback mixing on the recorder of recordings made of high SPL material, and then only because of the annoying quirk of Zoom's Mid/Side decoding implementation being pre-fader only.  Basically, that implementation eliminates the ability to adjust mix level of Mid/Side decoded channels upon playback, so all other channels must be mixed around the level of those channels as recorded.  If the omni channels then need a lot more fader level in the mix for instance, they can sometimes clip the mixer.  That doesn't always happen, but it can with very loud material.

I suspect that there is a gain setting (maybe no gain) that would give an identical result in either 24-bit or 32-bit float for the individual channels.

I think you are correct.  And it should if Zooms dual ADC auto ranging technique provides a wider dynamic range than that of the analog input front-end of the recorder.

[Ozpeter]

I think the whole thing can be viewed in reverse - if 32 bit float had been the norm for the last 50 years, and someone brought out a 24 bit recorder and said one of the features was that you can set the input level so that it doesn't overload or undercook, would people say, wow, that's a great idea?

Gutbucket, I'm not quite sure of the context but personally I was a heavy user of Voxengo MSED VST when remixing stuff in a DAW - that enables you to fiddle with mid / side levels to your heart's content with any stereo track, and of course is most appropriate if the source of the stereo track was an MS pair.  It basically re encodes the signal back to MS, then allows you to change the MS balance, then decodes back to XY.