The No.1 Website for Pro Audio
 All  This Thread  Reviews  Gear Database  Gear for sale     Latest  Trending
Quality Is Back - Keep Your High End
Old 1 week ago
  #31
Quote:
Originally Posted by RayHeath View Post
For a number of recordist, that is changing to 32-bit [wide-range] floating point over 48k or 96k even as we speak. Where it can be used, I suspect it will alleviate a number of unhappy accidents.


Ray H. [a.k.a. Dude]

Of course, the mix delivered to the client is generally not 32-bits.

Quote:
Originally Posted by Infa View Post
All your statements are totally true. I agree 100%. True too that nothing and no-one is ideal. Once you strive for perfection... oh man your in trouble, why ? Cause perfection doesn't and shouldn't exist. I do get all that even though I come across different. I just wish I seen people try like folks back in the day did. (this applies to literally everything)

These artist don't wanna get their knuckles bloody like they used to. They don't wanna *truly* sacrifice, don't *truly* invest in their goal. How can anyone expect a goal to pay off if you invest more in your car or clothes than you do your goal ? By the nature of law, your car and clothes will now bless you more than that said goal. LOL - Thats how things work.




Dude, totally correct. I just say 24bit, because FYI thats what it really is (hence the float). But of course I am recording in my DAW at 32bit float/ 96k all the time. Why not ?

And bros - Before the arguments start, LOL - lets just remember this thread isn't about the debate of 16 vs 24 vs 32f. I like the dynamic range headroom it gives me on drums, so lets leave it at that. There is a difference, just not how some think.

And 96k, I like the delivery smoothness and reverb tails better. And now - I can pass that directly onto Amazon ! Man, thats what this thread is about.

Quote:
Originally Posted by RayHeath View Post
Apologies for not previously being more explanatory. . .

As a long time pro software dev, I should likely make it clear that 'float' in my post refers to a floating point representation - as opposed to integer representation. Wide range refers to an encoding that maps those numerical representations to a very wide dynamic range. There are a number of 32-bit encodings for audio signals and they are not created equally.

In software, we use single-precision floating-point format largely to represent what most would think of as real numbers [and the like, kinda] - but complexity of this topic will rise very rapidly when one starts to ask a few simple questions. . .if one expects precise answers.


The common statement that "32-bit float *is* 24-bit audio" is no where near correct when discussing 32-bit wide range floating point encoding. In this encoding you wind up with somewhere on the order of 700 db increase [best I remember on the MixPre II implementation] in your available dynamic range - not a typo!

A primary consequence is that limiters are no longer required. . .hence no more unhappy accidents - e.g., from having the gain set too high.

Now, that's something to write home about!


Cheers,

Ray H.

Interestingly, most platforms I know of support this wide format - only be careful to turn down the volume on your earphones before hitting play.

[Edit] I forgot to sign using my 'Dude' moniker. Mmmm, next time.
I really don't see the point of storage at 32bit float....your capture is 24bit fixed, your DAW processes at 64bit float...what good does storing at 32bit float do you?

- it's not "8 bits better".
- there's 2 circumstances it could be useful - if you're doing a lot of offline rendering, and gain changes as part of that. Most people doing music won't do this. OR if (as you mention) you're not particularly careful with gain staging. Which is likely to cause problems down the line anyway (eg with setting compression thresholds) so it's not a great habit to get into.

Otherwise - I'm willing to be enlightened, but I simply don't see the point. Assuming one records at 24bit, doesn't process offline repeatedly and then mixes with competent gain staging?
Old 1 week ago
  #32
Gear Maniac
Quote:
Originally Posted by psycho_monkey View Post
I really don't see the point of storage at 32bit float....your capture is 24bit fixed, your DAW processes at 64bit float...what good does storing at 32bit float do you? [. . .]
Hi Adrian -

I expect for the work you are doing [as I imagine it] the value proposition isn't very strong. Making somewhat caricatured examples of your two circumstances:

1) Where dynamic ranges are simply too extreme. John Cage [Silence] vs Manowar [or Kiss] on Dueling Banjos?

2) Where the sound engineer simply cannot be local enough to the recorder to be 'careful' with gain staging. Film/video already have competing techniques, but this is not a small enhancement - and it will facilitate more creative work at lower cost.

In both cases you would later move likely both extremes toward the middle a bit for your mix. At some point you may not want to hold onto the 32-bit streams [a business decision].

A third scenario is where you need to send out a less capable sound tech to capture something and you want the closest thing to a guarantee there won't be any clipping. Or cases where you don't think something like Dugan Automixing will pull it off.

Yea, I think this is a really big deal! That extra 8-bits delivers an exponential-ish gain in the available safe dynamic range - a great bang for the buck!

Maybe not so much for your business model and workflows, though.

Although the format is not particularly new, SD's implementation of it is. And I noticed where Tentacle Sync was advertising it - but don't know much of their implementation yet. The Tentacle Sync should work with lavaliere and some DPA mics [I think] - but not with the larger 48v phantom powered mics [as the MixPre II recorders can]. I expect you will see more of this in the near future. Wish my Audio Limited A10s had it!


Best regards,

Ray H.
Old 1 week ago
  #33
Lives for gear
 

Ray, does this extra dynamic range headroom only happen on a/within a recorded file that was recorded at 32 bit float, or if recorded at, and session at 32 bit float is this extra dynamic range headroom also happening within your DAW's mixer too... like buses for example ?

I notice my buses seem to no longer clip into the red since I been recording at 32 bit float, they just go way past 0.0db yet are still only "yellow", not red clipping.

I didn't know if that was a direct cause of the 32bit float thing, or just PT new feature or something, lol Hadn't bothered to look not it. Plus, I'm scared who I ask. If you know what I mean.
Old 1 week ago
  #34
Gear Guru
 
Sounds Great's Avatar
 

Quote:
Originally Posted by Infa View Post

Quote:
Originally Posted by Sounds Great View Post
Now if they could only start streaming DSD.
Now your just making fun of me. lol Ok, I see how it is. Poke a stick at the funny guy.
Actually I wasn't joking. I haven't really heard that much difference with PCM as you move up from 16/44.1. But I do hear a major improvement over that with DSD recordings.
Old 1 week ago
  #35
Quote:
Originally Posted by RayHeath View Post
Hi Adrian -

I expect for the work you are doing [as I imagine it] the value proposition isn't very strong. Making somewhat caricatured examples of your two circumstances:

1) Where dynamic ranges are simply too extreme. John Cage [Silence] vs Manowar [or Kiss] on Dueling Banjos?
But you're still capturing at 24bit.

Quote:
Originally Posted by RayHeath View Post
2) Where the sound engineer simply cannot be local enough to the recorder to be 'careful' with gain staging. Film/video already have competing techniques, but this is not a small enhancement - and it will facilitate more creative work at lower cost.
But recording is still 24bit fixed too!

Quote:
Originally Posted by RayHeath View Post
In both cases you would later move likely both extremes toward the middle a bit for your mix. At some point you may not want to hold onto the 32-bit streams [a business decision].
But the DAW is working at 64bit float anyway!

Quote:
Originally Posted by RayHeath View Post
A third scenario is where you need to send out a less capable sound tech to capture something and you want the closest thing to a guarantee there won't be any clipping. Or cases where you don't think something like Dugan Automixing will pull it off.
Again - it doesn't matter what they store it as, it's being captured at 24bit fixed! if it clips on the way in, it's clipped.

Quote:
Yea, I think this is a really big deal! That extra 8-bits delivers an exponential-ish gain in the available safe dynamic range - a great bang for the buck!
Except as a storage format - it doesn't! I keep repeating myself, but the capture is 24bit, the processing is 64bit float. UNLESS you're processing files offline, 32bit storage isn't giving you anything.

Do you see what I'm saying?

Quote:
Originally Posted by Infa View Post
Ray, does this extra dynamic range headroom only happen on a/within a recorded file that was recorded at 32 bit float, or if recorded at, and session at 32 bit float is this extra dynamic range headroom also happening within your DAW's mixer too... like buses for example ?
Well, that's my point. you can't record a file at 32bit float - you can only store it like that! Your converters are 24bit fixed. There ARE 32bit converters, but given that a 24bit converter has a noise floor within that 24bits (ie you're not actually getting 24b of usable signal, I don't really see the point there either.

Quote:
I notice my buses seem to no longer clip into the red since I been recording at 32 bit float, they just go way past 0.0db yet are still only "yellow", not red clipping.
Clip your master buss and you'll hear it your converters are still 24b fixed. BUT your DAW is 64bit float - it's not the file storage that's causing you not to clip!

Quote:
I didn't know if that was a direct cause of the 32bit float thing, or just PT new feature or something, lol Hadn't bothered to look not it. Plus, I'm scared who I ask. If you know what I mean.
PT has been 64bit float since...PT10 I believe? or maybe 32bit float at PT10, 64bit float at PT12.

Last edited by psycho_monkey; 1 week ago at 07:03 AM..
Old 1 week ago
  #36
Gear Maniac
Quote:
Originally Posted by psycho_monkey View Post
But [. . .]
Hi Adrian -

You'll likely do better to watch a YouTube interview with with Paul Isaacs than to trust my memory. Here is one by Curtis Judd. But, I'll also respond and you can tell me how many I missed. I promise to treat this as a 'closed book' exam.

[1, 2] you're still capturing at 24bit. [. . .] But recording is still 24bit fixed too!

Well, no. The tracks are being captured at 32-bit wide range float - not 24 bit. I believe SD is using multiple analog-to-digital converters on each channel to pull this off.

An engineer is likely going to narrow each stream down. . .during [or early after] import into his/her DAW. I don't know what specific workflows each DAW vendor supports there.

[3] But the DAW is working at 64bit float anyway!

Yes, with exceptions [I suppose] that still use older 16/32-bit code or internally force smaller data streams. But for high-end DAWs, I hope this your assertion is correct.

While the DAW developer can widen either 24-bit integer or single precision float [32-bit] to a double [64-bit float], he or she can't get you back the lost information of the smaller dynamic range - only practically important in cases where it matters [e.g. , clipping].

[4] [. . .] if it clips on the way in, it's clipped.

See response to 1, 2 [above].

The predicate you're passing to 'if it clips' is false.

[5] [. . .] I keep repeating myself, but the capture is 24bit [. . .]

I'll let you do some research and/or watch a Paul Isaacs interview before you grade my test.

[6] [. . .] There ARE 32bit converters, but [. . .]

See response to 1, 2 [above].

To the best of my recollection, I think SD has a patent on their specific topology. I've not done any research to see how blocking their patent will be against other competitors.

[7] Clip your master buss and you'll hear it [. . .] it's not the file storage that's causing you not to clip!

Part 1. Agree. Note to readers [and self]: Don't clip your master buss.

Part 2. In competing architectures, the clipping [hopefully limiting, hopefully not overflow] would have occurred before storage. However, 32-bit floating point enables you to store a far larger dynamic range if you can find a way not to clip before persisting the stream(s). SD looks to have a reasonable implementation.


Again, I don't know that you are presently a target for any of this yet.


Best regards,

Ray H.
Old 1 week ago
  #37
Quote:
Originally Posted by RayHeath View Post
Hi Adrian -

You'll likely do better to watch a YouTube interview with with Paul Isaacs than to trust my memory. Here is one by Curtis Judd. But, I'll also respond and you can tell me how many I missed. I promise to treat this as a 'closed book' exam.

[1, 2] you're still capturing at 24bit. [. . .] But recording is still 24bit fixed too!

Well, no. The tracks are being captured at 32-bit wide range float - not 24 bit. I believe SD is using multiple analog-to-digital converters on each channel to pull this off.

An engineer is likely going to narrow each stream down. . .during [or early after] import into his/her DAW. I don't know what specific workflows each DAW vendor supports there.

[3] But the DAW is working at 64bit float anyway!

Yes, with exceptions [I suppose] that still use older 16/32-bit code or internally force smaller data streams. But for high-end DAWs, I hope this your assertion is correct.

While the DAW developer can widen either 24-bit integer or single precision float [32-bit] to a double [64-bit float], he or she can't get you back the lost information of the smaller dynamic range - only practically important in cases where it matters [e.g. , clipping].

[4] [. . .] if it clips on the way in, it's clipped.

See response to 1, 2 [above].

The predicate you're passing to 'if it clips' is false.

[5] [. . .] I keep repeating myself, but the capture is 24bit [. . .]

I'll let you do some research and/or watch a Paul Isaacs interview before you grade my test.

[6] [. . .] There ARE 32bit converters, but [. . .]

See response to 1, 2 [above].

To the best of my recollection, I think SD has a patent on their specific topology. I've not done any research to see how blocking their patent will be against other competitors.

[7] Clip your master buss and you'll hear it [. . .] it's not the file storage that's causing you not to clip!

Part 1. Agree. Note to readers [and self]: Don't clip your master buss.

Part 2. In competing architectures, the clipping [hopefully limiting, hopefully not overflow] would have occurred before storage. However, 32-bit floating point enables you to store a far larger dynamic range if you can find a way not to clip before persisting the stream(s). SD looks to have a reasonable implementation.


Again, I don't know that you are presently a target for any of this yet.


Best regards,

Ray H.
OK I see. You're referencing one particular box here! not even a DAW....it records 8 tracks direct to card. And yes, I can totally see why this might be useful for location sound.

BUT - none of this applies to DAWs. I could be wrong, but I THINK all current DAWs would capture incoming audio streams at 24bit fixed. Even if they store at 32bit. Even if using a 32bit float converter, unless there's some development I've not heard of. This may change of course.

To take advantage of this converter, you'd need to capture internally using it's 8 track setup, and transfer to DAW. Not the best workflow.

So I'll revise my statement - *unless* you're using this Sound Devices location recorder in 32bit mode and transferring to DAW, storing files at 32bit is pointless

I'm assuming the OP @ Infa doesn't have one, and isn't working like this - so my comments remain. 32bit storage captured by a 24bit recorder is pointless Unless you're doing a lot of offline processing, you're just using up space needlessly. It's a bit like "more is better" without really knowing why...

in fact, 32bit float capture in the studio is largely pointless - unless you're doing one take stuff live to camera or something (which is more of a post pro situation anyway). It's not hard to leave enough headroom! A lot of what you're saying about the advantages of 32bit float recording are to do with incompetency....

(and even with a 32bit converter - the idea clipping is "impossible" isn't quite true, you can still clip the mic pres!).
Old 1 week ago
  #38
Gear Maniac
Quote:
Originally Posted by psycho_monkey View Post
OK I see. You're referencing one particular box here! not even a DAW [. . .]
With respect to DAWs, here is a part of the interview you may find useful. . .Paul Isaacs on 32-bit and DAWs.

Here is another useful section about dynamic range where Paul eventually speaks to applications for sound designers, etc. A bit of his talk about mics input levels and associated voltages may also be kinda fun to listen to.

With respect to the 'one particular box', I don't currently know what the competition will be doing. But the functionality is so compelling, that I can't see them laying down - I'm expecting numerous vendors to head down this path. And I'm assuming the SD patent offers only limited protection from competition.

Time will tell. . .


Best - Ray H.

URL links above are provided for those who want a 'tl;dr' version of the interview.
Old 1 week ago
  #39
Lives for gear
Unfortunate the direction this thread took... the good news is that the lossless format is available at all. you can all argue until you're blue in the face about sample rates, but it's a fact that compressed audio files sound like crap. I'll take 44.1/16 uncompressed over ANYTHING lossy compressed, ANY day
Old 1 week ago
  #40
Gear Maniac
Quote:
Originally Posted by Infa View Post
Ray, does this extra [. . .]? [. . .] Plus, I'm scared who I ask. If you know what I mean.
Sorry, Infa for not giving a direct reply earlier. I expect your questions were all addressed in the Paul Isaacs interview and my chit-chat with Adrian. Please let me know if you've something outstanding.


Apologies to @ Philter that we didn't focus explicitly on bit-rate delivery [lossless vs lossy compression] from Amazon. I expect there is near unanimous appreciation of lossless here. And a big yes! from me.

But, I honestly don't know the answer to that one. It 'sounds' right to want lossless - but then people relatively rarely take the choice of lossless images/videos as consumers. I appreciate the difference - just don't know that much about consumer preferences. I'm thinking it will eventually be totally embraced.

The 32-bit float discussion is a very similar notion. . .but at the front-end capture and mixing stages. In certain circumstances it helps improve the 'garbage in' => 'garbage out' scenario. . .and strongly supports the idea that one will be delivering 24-bit 'lossless' audio to the consumer.


Best regards,

Ray Heath [Dude]
Old 1 week ago
  #41
Quote:
Originally Posted by RayHeath View Post
With respect to DAWs, here is a part of the interview you may find useful. . .Paul Isaacs on 32-bit and DAWs.

Here is another useful section about dynamic range where Paul eventually speaks to applications for sound designers, etc. A bit of his talk about mics input levels and associated voltages may also be kinda fun to listen to.

With respect to the 'one particular box', I don't currently know what the competition will be doing. But the functionality is so compelling, that I can't see them laying down - I'm expecting numerous vendors to head down this path. And I'm assuming the SD patent offers only limited protection from competition.

Time will tell. . .


Best - Ray H.

URL links above are provided for those who want a 'tl;dr' version of the interview.
I mean - the vast number of people using 32bit float recording are not using this box, therefore not getting the benefits you describe. In fact anyone using the format in a DAW isn’t getting the same benefit.

It’s not really about what happens in the future - I agree it’s a great concept for location recording. It’s that right now, people are using this format without a clue that it’s doing nothing for them.

And yes - not really anything to do with the topic, it’s just a personal bugbear that people see bigger numbers and think better without understanding the implications!
Old 1 week ago
  #42
Gear Maniac
Quote:
Originally Posted by psycho_monkey View Post
[. . .] In fact anyone using the format in a DAW isn’t getting the same benefit. [. . .]
I think I agree with everything in the post except this.
Old 1 week ago
  #43
Quote:
Originally Posted by RayHeath View Post
I think I agree with everything in the post except this.
Can you explain then? If you’re not capturing at 32bit float (which you’re not), if you’re not doing massive gain changes and rendering as new files, if your DAW is32bit float internal processing at least (which they all are) - what benefit do you get? Because I don’t see it!

Literally the only benefit I can think of is that if you’re internally capturing a mix, and you’re re-recording it to a new track, and you’ve clipped it. Storing it at 32bit let’s you fix that.

Or - you could just turn down your master fader before printing it to get the same effect.

What you DON’T get is any of the benefits - not one - that you listed above in relation to the SD capture device. You get no benefits in capture at all. So what’s the point?

Genuine question - I’m willing to be educated!
Old 1 week ago
  #44
Gear Maniac
Quote:
Originally Posted by psycho_monkey View Post
Can you explain then? [. . .]
Sure, and maybe we just talked past each other somehow. I'll try to be specific.

The statement - "In fact anyone using the format in a DAW isn’t getting the same benefit." - seems over-broad to me.

You later articulate [in a separate post referenced at the top of this one] two assumptions - [1] "If you’re not capturing at 32bit float (which you’re not), [and 2] if you’re not doing massive gain changes and rendering as new files," - I didn't see either of these constraints mentioned in the earlier post.

Now, these constraints in-and-of themselves don't have to be true.

If you elect to make them true, we're done here, and I agree - but only for that case.

As to the first assumption, one could very well have captured [or been handed] a 32-bit float file from a MixPre II.

As to the second assumption, again one could be handed a 32-bit float file created from who knows where. . .maybe generated via MATLAB or even by AI. The source is not important. The sound designer still wants it in the mix.

There are alternatives for sure, but importing them as 32-bit float into Pro Tools Ultimate would be my default path. They are coming into my DAW with the same primary benefits as previously discussed and referenced. [Footnote 1]


I'm a little concerned by the reference to internal processing width [let's pretend it is all 64-bit]. That width is often useful when I'm writing algorithms - as are higher 'sample' rates - to process the streams. But if something is missing. . .I can't get it back.


Hope this clarifies a bit. - Ray H.


Footnote 1: Well, almost exactly the same benefits. There was the scenario of a remix within the MixPre II - no extra functionality over a DAW aside from convenience for a quick hand-off. But I don't know if that is even possible if you also have to import streams from other sources. Bottom line: unless you've got someone breathing down your neck for a quick file dump, your DAW is going to be the default workflow for anything resembling a high-end job.
Old 1 week ago
  #45
Quote:
Originally Posted by RayHeath View Post
Sure, and maybe we just talked past each other somehow. I'll try to be specific.

The statement - "In fact anyone using the format in a DAW isn’t getting the same benefit." - seems over-broad to me.

You later articulate [in a separate post referenced at the top of this one] two assumptions - [1] "If you’re not capturing at 32bit float (which you’re not), [and 2] if you’re not doing massive gain changes and rendering as new files," - I didn't see either of these constraints mentioned in the earlier post.

Now, these constraints in-and-of themselves don't have to be true.

If you elect to make them true, we're done here, and I agree - but only for that case.

As to the first assumption, one could very well have captured [or been handed] 32-bit float tracks from a MixPre II.

As to the second assumption, again one could be handed a 32-bit float track created from who know where. . .maybe generated via MATALB or even by AI. The source is not important. The sound designer still wants it in the mix.

There are alternatives for sure, but importing them as 32-bit float into Pro Tools Ultimate would be my default path. They are coming into my DAW with the same primary benefits as previously discussed and referenced. [footnote 1]


I'm a little concerned by the reference to internal processing width [let's pretend it is all 64-bit]. That width is often useful when I'm writing algorithms - as are higher 'sample' rates - to process the streams. But if something is missing




Footnote 1: Well, almost exactly the same benefits. There was the scenario of a remix within the MixPre II - no extra functionality over a DAW aside from convenience for a quick hand-off. But I don't know if that is even possible if you also have to import streams from other sources. Bottom line: unless you've got someone breathing down your neck for a quick file dump, your DAW is going to be the default workflow for anything resembling a high-end job.
Right.

I'm possibly being over-broad, but not by much.

I don't think you're arguing with the statement that "for most people, storing files at 32bit offers no improvement over 24bit".

You're arguing a very specific set of circumstances. I don't even know how many of the Mixpre's are in the wild, but I'd be surprised if it's more than a couple of thousand. Most people won't ever see one!

I've already agreed that in some forms of post, it might be advantageous.

And yes - the processing depth of PT, and just about all modern DAWs I'm aware of is 64bit float. whether every process takes place at that depth I don't know - but my point is, you're not "gaining" processing depth during the mix by storing at 32bit float. It'll be exactly the same if you store at 24bit.

My point is really - there's an awful lot of musicians recording and mixing music through regular interfaces, never going near this mystical location recorder, storing their files at 32bit float, convinced they're working at higher quality.

In practice, it makes no difference other than using up 1/3 more storage space.

IF they're doing all the things you mention above, it conceivably might make a difference. Even then I'd argue in practice not so much, but it might at least!
Old 1 week ago
  #46
Gear Guru
 

Quote:
Originally Posted by RayHeath View Post
As to the first assumption, one could very well have captured [or been handed] a 32-bit float file from a MixPre II.
The vast majority of people on this forum probably don't even know what a MixPre II is.

If we were in the post production section then yes. Here, not so much.

Almost everyone is using 24-bit fixed as a target recording format as well as session interchange format.
Post Reply

Welcome to the Gearslutz Pro Audio Community!

Registration benefits include:
  • The ability to reply to and create new discussions
  • Access to members-only giveaways & competitions
  • Interact with VIP industry experts in our guest Q&As
  • Access to members-only sub forum discussions
  • Access to members-only Chat Room
  • Get INSTANT ACCESS to the world's best private pro audio Classifieds for only USD $20/year
  • Promote your eBay auctions and Reverb.com listings for free
  • Remove this message!
You need an account to post a reply. Create a username and password below and an account will be created and your post entered.


 
 
Slide to join now Processing…
Thread Tools
Search this Thread
Search this Thread:

Advanced Search
Forum Jump
Forum Jump