Jump to content

32-bit for Dialogue: Yay or Nay?


Marc Wielage

Recommended Posts

I was having a discussion with (what I suspect is) a young filmmaker on a different forum. and he's asking for a way to crank the post sound levels up more than 30dB, 30dB being the current limitation of Fairlight.

 

I asked, "why would you need to adjust post sound levels up +30dB? That sounds incredibly extreme!" I come from old-school Pro Tools where we were limited to I think either 10dB or 12dB without adding an extra Trim plug-in for additional range. Even 10dB for normal dialogue is exceptionally rare (to me). I did some quick math and came up to the conclusion that 30dB works out to 1000 times the intensity of a 0dB signal, (subjectively) 8 times as loud... which is a lot.

 

He responded that they had recorded all their dialogue in 32-bit, which would protect dynamic peaks and prevent clipping during the original recording. I responded that I felt that there's a point where self-noise in the microphone and in the preamp will be so overwhelmed by a 30dB boost, you're not going be able to take advantage of that range at all, and he got very defensive.

 

What nobody wants to come out and say is: my guess is they're trying to record sound without a location sound mixer on set to just adjust levels on the fly... as has been done for at least 80-90 years. My suspicion is they're operating on the belief that they can save money and just "fix it all in post," not cognizant that the sound mixer has to get things right on set, and they'll wind up spending way, way, way too much time tweaking the sound in post, wasting the additional time and money you thought you saved in post. And I think the noise floor will still be something they'll have to deal with.

 

What's the opinion here? Am I crazy that 32-bit recording is kind of a waste for regular (non-sound effects / non-music) location dialogue? To me, 24-bit digital dialogue has already got 144dB of dynamic range, and that's way, way, way more than can be reproduced in a theater. But maybe I'm thinking too conventionally, and maybe I'm too old-school.

Link to comment
Share on other sites

I totally agree. 32 bit is good for those who either don't have the time to learn proper gain staging or don't care. And for us, who can use 32 bit as a sort of backup system. There'll be some years now where people are trying to wrap their heads around 32 bit and how to get a good workflow, just like I imagine people did when the RED camera came out and people started having to use RAW footage in NLEs, and all the conforming ... I dunno. In a few years there'll probably be an AI that will level everything and take out any noise before the files are even imported in to a computer. Or even before it's recorded. 

Link to comment
Share on other sites

Well I think it’s a good question and worth revisiting. Definitely more products are coming to the market that are touting 32bit float as a feature. Zoom has a few items on the low end and then you also have Sound Devices getting into it with the mix pres and more interestingly with the A20 wireless. 
 

So Sound Devices now has a wireless transmitter that needs no gain adjustment on the pack which blows my mind and supposedly this works because of 32 bit… so if that all works as advertised I’d definitely conclude that 32 bit systems are very useful. 
 

I’m not sure mixers, recorders and the rest of the workflow is setup to take advantage of it yet.  I’d love to be able to turn in lower level ISOs without complaints from post. 
 

I think Zaxcom has done an admirable job tackling the frustrations involved with trying to record actors who whisper out one line and then blast the next one. The neverclip system works well with a strategy of giving you a LOT of headroom and then lets you manage both the iso track level and mix track level with various tools like adjustable compressors and variable iso attenuation. 
 

A goal for me is to never have to have mixes where I have one hand on a fader and one on a trim pot in order to control a really dynamic scene. Maybe 32 bit can help us get there?
 

 

Link to comment
Share on other sites

Just did a short movie where Post asked for the files to be in 32 bit. This didn't work with my cart setup of a Scorpio, CL-16 and a wireless cart with an A10-Rack on DANTE back to me. So I let Post know that no its still going to be 24 bit as that's all I could do. Then I remembered , the MixPre II-10 will record 24 bit and as I was using heavy wireless set up of 2 booms and body mics I could record a safety at the A10 Rack over copper to the MixPre 10 and run that as a 32 bit safety recording, whilst still keeping the Master files on the Scorpio. I ended up handing in two sets of files, A Master folder and a safety folder at the end of each day. It will be interesting to see what Post actually uses. My feeling is they'll stick with the 24bit Master files.

Link to comment
Share on other sites

FP32 is good for a OMB, though it will not prevent mic or the analog preamp from being over-driven. As was stated, no all NLEs (or DAWs) support floating point 32 bit, so that could be a post issue. Otherwise, If the audio files are separate (non-embedded in OMF or AAF), I would think TC would be maintained.

Link to comment
Share on other sites

[Note that this isn't written to enlighten anyone here (it won't), but for Marc to borrow or forward in his correspondence with the person who wants to add 30+dB gain. Not that he'll learn anything, either. I'm arrogant but not stupid.]

 

For dialog, I don't see an advantage to 32-bit over 24-bit with an OKish good limiter (and reasonable attention to levels). And good limiters are increasingly common. Also, beyond post hassles, the bigger downsides I see to 32-bit are that it gives people license to ignore gain settings....and inadvertently allows some to ignore good mic placement, with the thought that it can be fixed in post. I wonder if your correspondent wants to add 30dB+ to overcome suboptimal microphone placement.

 

I only have one 32-bit recorder, a Tentacle Sync Track E, and several 24-bit recorders (mostly Sound Devices but let's toss in my Canon C70 & Fujifilm X-T3). And as a mixer, I've fed all sorts of other cameras that record 24-bit audio. I've been handed a few tracks from other 32-bit recorders (MixPre II, Zoom F-series, other people's Track E recorders. I think that's it). But all of my "research" is at best anecdotal. 

 

In 32-bit mode, my Track E defaults to something like +18dB. If needed, I can crank up the recording gain with the smartphone app. The default seems to work fine for most dialog. The thing comes with a rather crappy lav (that looks a lot like what Sennheiser includes with their G3/G4 tx). The preamp & A/D are OKish (I mean for lower-end doc/corp and perhaps home-brew indie). I usually roll it at 24-bit (where the limiter is always on), but at the default level for 32-bit, a well-placed lav gives a decent signal. It certainly doesn't require an extra 30+dB gain.

 

I will guess that other 32-bit recorders also default to a setting that, when paired with a decent and decently placed mic, doesn't require 30dB+ gain in post.

 

And by way of comparison, I've rolled on dozens of interviews as an OMB with a well-placed Schoeps 641 going directly into my little Canon C70 camera at 24bit/48kHz. Those tracks sound much better than what I've recorded on the Track E (going through a SD mixer before my C70 sounds better, and recording to a SD mixer/recorder sounds a bit better still). And my Schoeps-->C70 tracks sound better than many of the 32-bit tracks I've been handed (and otherwise heard). But I more-or-less know what I'm doing. I think the meh 32-bit tracks I hear are related to things besides the recording format (mainly, not great mics and suboptimal placement).

 

I end up reminding people of (or teaching them about) the hierarchy of audio importance. As in the further to the left of this chain, the more it matters.

 

What's being said > mic placement > mic selection > wind/cable/handling/location-noise mitigation > recorder/preamp choice > gain staging > recording format.  

 

None of that is news to anyone here, though there's room for more and for debate. And sure there are a range of quality issues and craft at each stage. But for someone who wants to add 30dB to a dialog track in post, I think this is enough to get them to pay increased attention to the first four steps, while still paying enough attention to gain and format to keep themselves (and their post pals) out of deep trouble. So then they don't need 32-bit audio for dialog. Or at least, they won't need to add 30dB gain in post to get to a reasonable level.

 

So my point is, good 24-bit w/limiter audio is widely available and works well (and is basically simple). 32-bit works, but isn't a panacea that lets you disregard other more-important issues for capturing good dialog.

 

Disagree with what I wrote? That's fine. I'm still recovering from Covid-19, so that's my excuse. 😉

 

Jim

PS- I'm fine, btw. Mainly now just a bit of a cough and tiredness. 

Edited by Jim Feeley
Corrected spelling of Marc's name. Sorry Marc!
Link to comment
Share on other sites

I'm responding to Jim just above and Marc at the top so apologies if I'm repeating anyone by accident. The question really is - "one or two cues?" - ie trying to save the day, or has everything been recorded 'off mic' ?

 

Sadly there have been several technological shifts and advances which have been misunderstood, 32 (or even 24) bit being one, ambisonics being another, and at the end of the day it hasn't been (hopefully) our fault, as the experts, nor even the (generally disappointed) user/producer's fault, who have just been told something. It has always been the over enthusiastic marketing or promotion of an application, technique or (cheapness-driven) improvement in standard which has been the culprit.

 

Perhaps it is time for us (and technicians in other fields) to be more vocal, less supportive but more explaining of slight shifts when they arrive. A case in point: 16 bit was decided as the first standard (at least in our field) for bit depth when Nagra and Akai and Sony etc etc were producing 18 and 20 bit products. Rather than have a new 'next standard' of 20 bit, 24 bit was decided upon. A good choice in my opinion, even though for years the actual converters providing a 24 bit end result were not necessarily 24 bit.

 

And consoles have been beyond the 'new consumer heaven' of 32 bit for as long as I've been pushing P&Gs.

 

Jez

Link to comment
Share on other sites

I'd be fine with us all moving to 32bit files for delivery IF the post industry adopted proper workflows for it AND the major NLE's supported exporting files out in the vein of OMF/AAF that could handle it.  I do a little post for short docs and someone recorded theirs in 32bit, edited it in 32bit in Premiere, and then when it came time to exporting to me for Pro Tools, we had to do a whole song and dance of using a combination of Adobe Audition and AATranslator (a great tool) to make it work in a way that then could be spit out to Pro Tools in 32bit.  Was very time consuming and we had ten episodes to convert over so it was days of just getting the files ready vs hitting export an AAF and spit out a Quicktime per usual with 24bit workflows.

Link to comment
Share on other sites

1 hour ago, Derek H said:

32 bit float can record 770dB above 0dBFS

Which is great for all those times we have to record an ant fart, then one second later the space shuttle takes off - and there is literally no time to adjust the trim.

 

And it finally lets you take full advantage of that 1000dB dynamic range microphone. 

Link to comment
Share on other sites

4 minutes ago, Derek H said:

OK. Honestly, it does seem like that’s what I’m trying to record some days. 
 

I’ll bow out of this discussion for now. Maybe I’ll check back in 5 years and see what we’re all recording on at that point. 

Sorry Derek, I see how my post may seem like a snarky comment, but it was more an attempt to be humorous. I definitely didn’t mean to shut you down or have you bow out of the discussion. We’re all here to exchange viewpoints and yours are certainly as valid as anyone else’s!

 

And, I do see some scenarios where 32-bit float might come in handy, but it probably (hopefully) will not replace the services of a skilled sound mixer. It may help salvage some things, but it also takes some extra time/work/$$ for post to actually use it.

 

Maybe in 5 years all major NLE will be able to import and and take full advantage of 32-bit float files, but for me the question remains; is it necessary?

Link to comment
Share on other sites

Thanks Johnny, it’s all good. I’m definitely as guilty of snarky posts as anyone. For me I’ve spent the last couple years ignoring 32 bit float thinking it’s the same as 24 but with additional (mostly unneeded, dynamic range) but when I realized that it’s really very different from the standard fixed point method we’ve all been using and seeing it used in the new A20 wireless has peaked my interest in it. That and finding out you can record above 0dBFS and still recover it in post. 
 

I have no interest in enabling amateurs only in having better tools and methods for professional use. 


Side question… anyone know what is the technical reason DAWs and NLEs have trim gain limitations? Just coded that way out of habit or is there an actual reason they can’t boost more than 10, 18, or 30dB?

Link to comment
Share on other sites

Interesting thread. Mostly what I’m hearing is that 32 bit float is just fine if your sound person is skilled at his craft -things like mic placement, boom handling, eliminating motor noise, etc, AND the post workflow has been tested and verified. Right now the latter is bleeding edge so invites issues. But hopefully NLEs and DAWs will catch up soon.

 

 As an indie editor, one thing I sometimes run into is a novice field mixer riding gains inexpertly, resulting in awkward level changes in the middle of words and requiring me to “undo” in the rough mix their frequent gain changes. This makes me hope I’ll start seeing 32 bit float so that’s one issue I’ll no longer have to deal with. 
 

Cheers!

   Joe
 

 

Link to comment
Share on other sites

6 minutes ago, Derek H said:

Side question… anyone know what is the technical reason DAWs and NLEs have trim gain limitations? Just coded that way out of habit or is there an actual reason they can’t boost more than 10, 18, or 30dB?

 

I can't say I 'know', as I'm just an end user, not a manufacturer, but I would say:

 

1. Common sense: Health and Safety - so my ear drums (and, of lesser importance, monitors) aren't blown away by the idiot wanting to play me something

 

2. Losing years worth of good workflow practice of proper gain staging.

 

In sound NLEs at least there should always be a way of cranking gain up beyond the chosen limit (normalise a copied clip to a chosen threshold for instance) which is preferable to something which can be done easily either in ignorance or by mistake.

 

Jez

Link to comment
Share on other sites

1 hour ago, Derek H said:

Side question… anyone know what is the technical reason DAWs and NLEs have trim gain limitations? Just coded that way out of habit or is there an actual reason they can’t boost more than 10, 18, or 30dB?

 

My best guess about this is that most DAW's are loosely based on consoles and console workflow which had minimal amount of boost that could be applied to each channel and infinite attenuation.
There is no technical limit, just a programmed one.

To answer the OP's other discussion, if they need to add 30dB of gain they should process the clips with a gain plugin rather than try to add it at the track level.  That's assuming I understand what and how they are trying to accomplish.

Link to comment
Share on other sites

Also, am I the only one who dislikes using the phrase "record above 0dBFS"?  0dBFS is FULL SCALE, meaning the maximum amplitude a digital system can represent.  So without qualifying it as specifically 0dBFS relative to a 24 bit system, that phrase doesn't really have a specific meaning?  i.e. a 32 bit system has a different FULL SCALE than a 24 bit system, and although their maximum amplitudes are drastically different, they are both referred to as 0dBFS.  I guess I could go around saying a 24 bit system can record 48 dB above 0dBFS of a 16 bit system, but why would I?  It's a confusing statement.
Instead speaking in terms of dynamic range is more accurate without needing qualification and better describes what we are trying to highlight imo.  But then again, if we do that, how can we promote that new technology is accomplishing the impossible task of recording a digital signal above FULL SCALE?

Link to comment
Share on other sites

2 hours ago, Wandering Ear said:

am I the only one who dislikes using the phrase "record above 0dBFS"?

No, you're not.  I find it incredibly misleading as well, as it replaces a coherent technical scale with, essentially, a conventional scale.  An unfortunate piece of marketing genius to convince people that it is superior.

In reality, the biggest difference actually has to do with the fact that it's floating point rather than integer, and that has ramifications for where in the scale it is best to record.  With floating point, precision (and therefore audio resolution) is highest in the *middle* of the scale, whereas integer formats are equally precise throughout the scale, but there is a quantization noise floor at the bottom of the scale, which means the "best" audio is at the top end of the scale (i.e. "full scale").

It's been a little while, but if I recall correctly, the difference between the maximum value that can be represented in 32-bit float and the second-highest value is something like 8dB, which would produce very, very bad results if we tried to record at the top of the scale the way we do with integer formats.

In reality the precision of 32-bit float is identical to 24-bit float.  32-bit float uses 23 bits for the base, plus 1 sign bit = 24 significant digits.  The remaining 8-bits are an exponent (mantissa), which is what creates the massive 770dB increase in dynamic range ... but at the cost of a loss of precision at the outer bounds of the scale.

I think it would be more accurate simply to stop using dBFS as a scale for floating point and recognize that 32-bit float requires using a different reference point for best results.  We need something like dBFP (dB floating point), where convention dictates that 0dBFP is the middle of the floating point scale (i.e. 0.0), and 72dBFP = 0dBFS.  If I'm doing my math correctly, this convention would properly match 32-bit float and 24-bit int such that both formats would be used to their highest potential at similar recording levels.

Presumably, a similar convention is already being used under the hood by most DAWs, and the claim of "exceeding" dbFS 0 simply indicates when signal levels exceed what a 24-bit int format would be capable of representing (i.e. the point at which precision starts to degrade in floating point).

A major disadvantage of switching scales is we'd all have to adapt to new standards for "correct" recording levels.  Presumably, we'd all have to target somewhere around 52dBFP for "normal" dialogue levels, and I can only imagine the amount confusion this is likely to cause between set and post...

Link to comment
Share on other sites

4 hours ago, Wandering Ear said:

Also, am I the only one who dislikes using the phrase "record above 0dBFS"?  0dBFS is FULL SCALE, meaning the maximum amplitude a digital system can represent.  So without qualifying it as specifically 0dBFS relative to a 24 bit system...

 

I think it's reasonably helpful to make clear(ish) that at 0, 32-bit files match with 24-bit and 16-bit files. Perhaps after some time has passed, we can shift to a scale nomenclature that reflects the true value of 32-bit audio. Such as:

 

Decibels Beyond Scale, or dBBS

 

I jest. Also, Jez, I may not have the hours of many here, but I've been in and out of professional audio for several decades now and have been through a bunch of technology shifts (as have we all). Part of the issue with 32-bit adoption is it's happening through people who have less contact with craftspeople and technicians than happened in the past. And as (I think) you say, some view 32-bit as a panacea for all audio ills. 

 

I'm still wondering why Marc's correspondent wants to add over 30dB to his track. Did he intentionally lower the gain setting on his device from the default, and/or was his mic poorly placed? Whispering talent? Or...? 

 

(Marc, to be clear,  I'm not asking you to follow up with the person. And sorry about misspelling your name earlier.)

Link to comment
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

×
×
  • Create New...