Topic: Zoom F6 (32-bit float equipped)

[Walstib62]

ok then list the "incorrect statements" ive made. im not aware of any concert tapers that are recording concerts peaking at -60dB for shits and giggles, which again means that 32bit float offers *zero* advantage compared with properly set levels ( peaks > -20 dB) for the purpose of concert taping

Very telling that the subject had literally sat for close to a year, until , just coincidentally after you develop a habit of trolling me in the PZ board (The majority of your posts in the last 10 days have been directed at me, into double digits), you feel the need to follow me over here and "correct the record".
 
so make with your highfalutin' electronics knowledge, then. No need to be coy and make there be (7 so far) extra posts. There is still time to save a bunch of people from misleading info, instead of wasting your time making the majority of your posts on political issues. I'm sure we'd all be happy to hear about your hands-on experience with the F6.

Ok.
1. I don't have any hands on experience with the F6. I have, however a lot of experience recording in 32 bit using a USB interface to PC, storing directly to HDD via eSata. Here's one example of many such recordings I have done. This goes back about 9 years, so this concept is not new to me. Bear in mind, this is a cheap interface, nothing fancy by any means.

  https://archive.org/details/donna2011-10-09

2. Your blanket statement that there is "zero advantage" to recording in 32 bit over 24 bit is incorrect. It may be true that you can record in 24 bit with the Peak amplitude not going over -20dB, as you say. If this is your go-to approach, and you like it, then more power to you. Ideally, I want the average signal amplitude to be at -18dB. This would generally provide for enough dynamic range (quantization in the ADC, which is what we are talking about here) at the lower end of the overall signal amplitude with enough headroom at the top to prevent clipping. However, let's say that you are recording an artist that has a much wider range in dynamics (SPL as presented to the mic, for instance). If you recording in 24 bit, you can absolutely clip the signal at the recorder before the mic reaches its' maximum SPL limit. This is not the case with 32 bit recording with a properly implemented input gain stage. In this scenario the mic will clip before the recording stage clips. Of course, this can be prevented in 24 bit by riding the recorder gain and tweeking up or down, but who would do that if it is not necessary? I find this to be a huge advantage in field recording.
 On the other hand, if what you are recording has a more limited dynamic range, then the advantages in 32 bit are less evident. This I will agree with. However, the blanket statement that there is zero advantage to 32 bit is incorrect.
 

[Gutbucket]

While composing on this one the two posts above appeared.  I'm hitting post anyway with the hope this may clarify things for carpa.

The increased dynamic range capability of the recorder without need for making a manual adjustment is real. It works. That was achieved with some new innovative digital techniques (call it fuckery if you like.. in some sense, all forms of digital recording are that).  As far as I'm aware the scheme seems audibly transparent, based on the reports of users and and the recordings I've heard made with it. 

At the same time, the need for such increased dynamic range without requiring a manual adjustment has not been particularly well established with the majority of live concert taping done by members of this forum.  And with it come some associated burdens.

^ This is the situation on the face of it.

There are times when not needing to make any adjustment will save a recording, yet after one becomes familiar with the system they are using those times are rare and generally when recording something other than what one normally records.  It is not particularly difficult for recordists to comfortably fit the dynamic range of a recorder to that of the music.  The learning curve for doing that is not steep.  It mostly takes doing it a few times to figure out.  This is especially the case if you tend to record the same type of music in pretty much the same situations each time - such as recording oneself playing piano. 

All that is somewhat trickier if one is unable to tend to the recorder while recording, which increases the value of the freedom from making input level adjustments (especially the first few times as a novice recordist).  But relief from the need to adjust input levels brings with it associated burdens- file sizes are larger, software which can handle the more obscure recording format is limited, and the need to process the recording afterwards for distribution or simple direct listening is increased.  Those things have been discussed in these threads, but have not really been clearly laid out as the "cost burden" of using 32-bit floating point recording formats.

[voltronic]

Gut,

Thanks for weighing in with some clarity.  I agree with everything, except I would say the "costs" of  32-bit FP have been rather small from my perspective.

1) The file size increase from my previous go-to format of 24/48 to 32fp/48 is about 30%.  That's not nothing, but with storage being so cheap nowadays, it has made almost no impact.  I always compress my masters as FLACs and delete the original WAVs after I have finished my post work, so then the long-term storage difference is even less.

2) Regarding software: I think the list of audio editing software that doesn't handle 32-bit FP files is rather small today.  It used to be more like DSD, where you only had a few choices.  As far as playback, then yes you are more limited unless you convert to a fixed-point format.

[Gutbucket]

FWIW, I'm not suggesting that the increased burdens aren't worthwhile, only that they should be clearly acknowledged and have yet to be as concisely defined as the benefits.  Hopefully this dialog helps to clear this part up a bit.

I've a few practical questions (IME, most of the value-important choices with regards to taping hinge upon practicalities rather than absolutes)

1) Can the raw recorded files be played back easily directly off the recorder?

2) Can the raw recorded 32-bit files be FLAC'd for storage without having to do any processing other than renaming the files if desired and doing the FLAC compression?

3) Can the raw recorded files (in their native file format and/or FLAC compressed format) be played easily via standard computer media players without having to undergo file conversion?

^ Lack of these capabilities would represent a significant burden for my usage, file maintenance, and workflow.  I acknowledge lack of those capabilities may not significantly impact others in the same way.

FWIW, in one way the Zoom F6 may be somewhat less burdened in this regard than the SoundDevices recorders, especially since SD has yet to include 32bit FP support in their WaveAgent app. I think F6 can be set to save 32-bit floating point output files as multiple individual mono or multiple 2-ch stereo files rather than only as a single multichannel polyfile.  This is partly my assumption from use of Zoom F8 and is important to me for file management by eliminating most file-size related file-splits and by allowing direct listening to the files in a software player without having to load them into an editor.

There is nothing here that would stop me from recording and enjoying the benefit of never having to adjust recording levels, yet these things represent significant impacts on convenience and the way I work and will influence any decision I make about using 32-bit FP recording.  I'll also note that none of the current 32-bit FP capable recorders require recording in that format, so the mere presence of this capability needn't influence one's choice of recorder if one is not planning to use it for whatever reason.

[carpa]

Very interesting thread, I can say. Last Gutbucket's question are really interesting and I'll also read the answers with attention. Yet, assuming the burden of a raw recording to be adjusted in post could lead to a bigger work, I'd say that the 32 bit recording could even play in case  a safety file if a dual 24 bit recording has been made, as I guess this f6 can do.
Should the 24 bit file ( with level setting allowed) work perfectly fine, great; you could eventually decide to take this one and cancel the bigger 32 raw file.
If, on the opposite , the 24 bit results too conservative so to need post adjustment work or - worse- should result as clipped, then I think we'll should welcome the extra work on the 32 bit file.
Something like the capability most recorder have to track a "parallel" recording 12 db less or so, but with more quality, I assume.

[voltronic]

1) Can the raw recorded files be played back easily directly off the recorder?

Yes, with a slight caveat: As previously stated, input gain becomes fixed in 32-bit FP mode.  You can choose (as I have) to set the track knobs to control the post-ADC record level, rather than faders for the L/R downmix (which I never use).  If you set those levels high enough so that your resulting file has levels over 0 dBFS, you will clip the headphone out because the DAC is just playing back what it's being fed, which may be a file with clipping, even though it is fully recoverable in post.  If your levels are set more conservatively so that you don't have any overs, that won't happen. 

2) Can the raw recorded 32-bit files be FLAC'd for storage without having to do any processing other than renaming the files if desired and doing the FLAC compression?

Yes.  That's exactly what I do.  I have a Converter preset in foobar2000 called "FLAC, no changes" that preserves bit depth and sample rate, and just encodes to FLAC level 8.  The result is still 32-bit float point.

3) Can the raw recorded files (in their native file format and/or FLAC compressed format) be played easily via standard computer media players without having to undergo file conversion?

Yes.  I use foobar2000 as my audio player of choice, but I just tested and Windows Media Player plays them just fine.  That's about as generic a player there is.

FWIW, in one way the Zoom F6 may be somewhat less burdened in this regard than the SoundDevices recorders, especially since SD has yet to include 32bit FP support in their WaveAgent app. I think F6 can be set to save 32-bit floating point output files as multiple individual mono or multiple 2-ch stereo files rather than only as a single multichannel polyfile.  This is partly my assumption from use of Zoom F8 and is important to me for file management by eliminating most file-size related file-splits and by allowing direct listening to the files in a software player without having to load them into an editor.

If I was a SD customer, I would be very frustrated by that, especially with their own software not being updated to support the file formats their newest recorders create.  The F6 does indeed give you more options as far as recording formats go.

On the flip side, the MixPre II units will allow 32-bit float recording in interface mode (on OSX only); whereas the F6 is limited to 24/96 fixed.

[Gutbucket]

Thanks volt, that puts a few of my previously outstanding practicality-related concerns to rest.

[aaronji]

There are also differences in implementation between the SD and Zoom flavors. I have no idea to what extent that influences sound quality, but SD obviously thought it was worth a lot of R&D and a patent application. As I understand the patent, it is a little bit hardware but mostly it is software to deal with the issue DSatz has described (admittedly based on a cursory reading and when it was first published). Perhaps that offsets the inconvenience of the polywave format (although, personally, I don't find that to be a problem).

[jerryfreak]

whew quite a bit to catch up here, excuse the long combined post

On the other hand, if what you are recording has a more limited dynamic range, then the advantages in 32 bit are less evident. This I will agree with. However, the blanket statement that there is zero advantage to 32 bit is incorrect.

we are actually in agreement, that is *exactly* what i said. Based on typical use case of most tapers here, very limited (more like no) advantage for most. I never said zero advantage for all applications and i trust in Voltronics extensive use in his application.

I'm not sure i quite understand the applicability of dumping a 24-bit USB signal into a 32 bit container on the computer to discussion of this gear. Nothing is gained in that scenario other than limiting quantization errors on downstream processing, which is basically how every wave editor already works fundamentally (regardless of native bitrate, all calcs done in 32bitFP, and a final render/dither to bitrate of choice).

Its fundamentally different than the application we are discussing here, the architecture of the 32bit FP recorders which use multiple ADCs to overlap in proprietary ways to provide a larger ADC dynamic range

It may be true that you can record in 24 bit with the Peak amplitude not going over -20dB

in a live concert environment with a recorder with a similar 120dB+ EIN, probably more like -40dB or more. Ill do some tests and well see if people can pick them out

However, let's say that you are recording an artist that has a much wider range in dynamics (SPL as presented to the mic, for instance). If you recording in 24 bit, you can absolutely clip the signal at the recorder before the mic reaches its' maximum SPL limit.

of course you can. You can also clip the same input at the same SPL at 24 bit. The zoom has relatively modest max input level 0f +24 dBu in line-in mode (relative to some other gear that can take signals approaching +30 dBu)

Whether you record at 24 or 32FP, that max input level does not change

In regard to recording at +24 dBu line in, vs +4dBu mic in, all the zoom is doing in the former case is using a 20 dB pad. Ive tested it myself that recording with this pad (aka "line-in) raises the very real noise floor of the analog input substantially

At the end of the day there is no free lunch. The analog input is still constrained by its max input level and EIN. You cant entirely scrap the analog gain stage like in this design, and rely on wide ranging ADs and not have compromises. This is why real preamps that can offer 60-70 dB of clean gain are so valued. Im sure that in theory, that with much more expensive/complex gear, there could be an input stage that offers more substantial use of the 32-bit FP but youd be hard pressed to find an application to take advantage of it (recording bird calls next to a rocket launchpad, perhaps, with some advanced imaginary transducer with over 130 dB dynamic range)

If you recording in 24 bit, you can absolutely clip the signal at the recorder before the mic reaches its' maximum SPL limit. This is not the case with 32 bit recording with a properly implemented input gain stage. In this scenario the mic will clip before the recording stage clips. Of course, this can be prevented in 24 bit by riding the recorder gain and tweeking up or down

stepping back a bit this is demonstrably false. again its the same analog input regardless of bitrate/ even in 24-bit fixed the dynamic range of the input does not change. And you dont need to be anywhere close to 0dB with 24 bit, its still relatively easy to leave yourself, 12,20, possibly 36 or more dB of headroom. certainly no need to 'ride the levels'. Again ive successully abused good 24bit ADCs, eschewing a preamp, and made concert recordings peaking near -30dB with the noise floor inaudible beyond ambient noise.

2) Can the raw recorded 32-bit files be FLAC'd for storage without having to do any processing other than renaming the files if desired and doing the FLAC compression?

i record in polywav, and end up saving my raw 32-bit polywavs as large .w64 files from soundforge

FLAC can do 32FP stereo files fine,  I think fundamentally the standard allows FLAC to handle polywavs, but my standard workflow for making FLAC files chokes on it somewhere, so i just do as above. the .w64 files are relatively enormous, >10GB per set at 6 ch of 32/96

3) Can the raw recorded files (in their native file format and/or FLAC compressed format) be played easily via standard computer media players without having to undergo file conversion?

foobar plays back  2-ch .wav and .flac fine

I'll also note that none of the current 32-bit FP capable recorders require recording in that format, so the mere presence of this capability needn't influence one's choice of recorder if one is not planning to use it for whatever reason.

cheers to this. even in absence of having the 32-bit feature, i still think the Zoom F6 is an outstanding piece of hardware for the price. In fact, some of its better features (like fader adjustment and simultaneous USB output) are crippled in 32FP mode

There are also differences in implementation between the SD and Zoom flavors. I have no idea to what extent that influences sound quality, but SD obviously thought it was worth a lot of R&D and a patent application. As I understand the patent, it is a little bit hardware but mostly it is software to deal with the issue DSatz has described (admittedly based on a cursory reading and when it was first published). Perhaps that offsets the inconvenience of the polywave format (although, personally, I don't find that to be a problem).

i think their trick is using 3 ADCs instead of 2, and interpolating the handoff between them somehow. Again I dont think that weve seen that in either design that handing off between multiple ADCs is causing distortion in either device

I'm not sure that their design makes any better use of the real-world constrained dynamic range of analog input in portable gear

i think we all win in the end when manufacturers take their own path to achieve their means, healthy competition breeds innovation.

[aaronji]

think their trick is using 3 ADCs instead of 2, and interpolating the handoff between them somehow. Again I dont think that weve seen that in either design that handing off between multiple ADCs is causing distortion in either device

I also think they are using 3 ADCs and then using regression over a (relatively) small number of samples to approximate the instantaneous rate of change to facilitate the "handoff". Both of which could potentially reduce noise floor changes. Of course, I might be entirely off-base; like I said I read it over pretty quickly and it was a while ago (and it's not my field)...

[jerryfreak]

sounds like it could be a different implementation of the "look ahead limiter" philosophy used in a lot of gear

[voltronic]

cheers to this. even in absence of having the 32-bit feature, i still think the Zoom F6 is an outstanding piece of hardware for the price. In fact, some of its better features (like fader adjustment and simultaneous USB output) are crippled in 32FP mode

Fader adjustment is definitely possible in 32-bit float mode; it is only preamp gain that is fixed in that mode.  You actually have a choice on the F6 of how those faders behave, or more accurately, where in the signal chain they take effect, by going under under Menu > Input > Track Knob. 

Hear me out here:

If set to Reference Level, the track knobs are faders for the L/R downmix.  I never use the downmix, so I prefer setting to Rec Level, which actually changes the levels of the 6 ISO tracks being written to the card, post-ADC.  So even though the gain is fixed, you can set to Rec Level and crank the hell out of your channels so you have ridiculously high (or low) levels being written to the files.  In Rec Level mode, the downmix faders are bypassed, but I still consider the knobs to be faders because their level adjustment is happening well after the input trim (again, which is fixed in this mode).  See pp. 194 and 198 in the manual which show how this works.

I have not yet tried using my F6 in interface mode, so I cannot speak directly to the USB output.  I know it is possible to record to both the SD card and also output through USB at the same time, but the manual makes no mention of 32-bit float being unavailable in this mode.  I know that 32-bit float cannot be output over USB, which I hope might be changed in a firmware update.

[jerryfreak]

yeah that shoulda said 'trim' not 'fader'

its been awhile but im still fairly certain based on block diagram and convos with zoom support techs, that the F6 has **no** analog gain adjustment, in either mode, its 'gain' is limited to the 'trim' feature which is entirely in the digital domain, which would (in my mind/use case) kind of make it a fader by definition. while "trim" is fixed in one mode, thats kind of a ruse as it technically doesnt exist (pre-ADC) in the other mode either. i believe that in that mode all its really doing is fixing the exponent in the 32-bit isos, i could see where this would be visually more challenging/counterintuitive to someone used to conventional methods, but in the end the result should be equivalent


Of course thats more of a use case/workflow issue, i personally only use ISOs and dont use downmix tracks or outputs. In either case, it would be the same though, as despite the isos being at their own individual level, you still have that ability to downmix them at individual levels


at the end of the day, the best performance is going to be achieved by matching input device to clip at the + 4dB (mic-in)/+24 dB (line-in) level of the device (Technically the same but in theory the former would be superior if you want to eliminate a pad from the signal chain)


as of now the unit cannot operate in usb mode while in 32-bit mode (or with sample rates over 48K). Back in January i requested zoom address this in future firmware if possible, got no reply on whether it would be considered or is technically possible

[Paul Isaacs]

Hope you don't mind me gate-crashing this thread.... full disclosure - I work for Sound Devices.

I just wanted to clarify some facts about the Sound Devices MixPre-II 32-bit float implementation that have come up in this thread.
But first, I would like to say that both the F6 and MixPre-II are viable machines if you are looking to do 32-bit float recording. They implement said technology in different ways from not just an internal hardware/architectural point of view but also from the user experience point of view. That said ...

1) MixPre-II mic preamp can handle a maximum input mic signal before clipping of +14dBu, that is 10dB more than the F6.
2) MixPre-II HP output is driven by 32-bit float to analog conversion, thus there is no clipping either during playback or recording when signal exceeds 0dBFS. Be careful as the HP out can get exceedingly loud  with signal >0dBFS. Turn the HP vol down to protect your eardrums and reduce the likelihood of HP amp itself clipping.
3) At the time of me posting this, the MixPre-II is the only USB audio interface which can stream 32-bit float audio to a computer, both MAC (Catalina and higher) and Windows (10 and higher - you'll need the special ASIO driver on our website to enable it).
4) MixPre-II can simultaneously record 32-bit float and be a USB 96k/32-bit float multichannel interface to/from your computer.
5) MixPre-II can, during playback, remix/re-record 32-bit float recordings that exceed 0dBFS to below 0dBFS. You adjust the fader knobs, pans, mutes etc  on playback to achieve the new mix. The result is a 32-bit float 1ch or 2ch wav file.
6) The WaveAgent app does not currently support 32-bit float wav files. If you need to convert 32-bit float polyphonic wav files to monophonic, there are many apps out there that can do that.
7) 32-bit float wav import and 32-bit float USB audio streaming are quite well supported in DAWs now. Even the MAC Finder will play them back. This might be useful: https://www.sounddevices.com/32-bit-float-applications/

Paul, SD

[voltronic]

Thanks for weighing in, Paul.  I wish we had a rep from Zoom here to answer some specific questions I have about their implementation - I think I'll be sending them an email soon. 

Regarding the input level: Clearly MixPre-II has a distinct advantage.  It's worth noting that the F6 allows phantom powering while the inputs are set to LINE, and in that mode they can handle an input level of +24 dBu.  I haven't tested it myself, but I imagine that a pad is being inserted in that mode because the impedance increases by 2 kOhm as well.

I need to test the HP output of the F6 further to see if the behavior there is different than what you are describing for the MixPre-II.  When I mentioned clipping before, I was thinking about the HP amp.