Cam Audio Recording Request: 32-Bit Float

I don't understand why did this thread was turned into an anti-32bit float campaign.


Tim made a reasonable request, and while it may be unlikely in the near future, cameras - especially those aimed at content creators - may at some point offer 32 bit float recording. For it's obvious workflow advantages for certain target groups.

Not that a single camera offers this feature

Is that so? It does involve an optional hardware feature ($500), but this is in-camera 32bit float audio recording. And doubtless the first of many. Perhaps you need to consult more youtubes?

Or that he’s shown a single example of him using 32-bit float. In fact, so far I’m the only person in this thread who’s discussed an actual case of using 32-bit float and why I used it, and shown the related waveform.

I wasn't aware it was necessary to provide examples. Meanwhile, your confidence in the value of your contributions -- what was it again? a waveform which purports to be a thunderclap, demonstrating what everyone already knows? -- just might exceed the general assessment.

For a guy not looking for a fight, you sure are generous with the provocations.....

"The loudest known sound in history was the eruption of the Krakatoa volcano on August 27, 1883, which was estimated to be 310 decibels (dB)."

32 bit float would handle that, of course you and your mics would be cinders.

Good Luck

I'll be honest, 20% Voice Isolation in Resolve easily gets rid of any air conditioner units. To such a degree that I don't even worry about them anymore knowing it's that easy to get rid of them. Still, if I can turn it off I will for the cleanest possible audio. Yet they're not really a worry these days.

I should add Noise Assist on SD recorder/mixers can be non-destructive too; you can set it to the busses only also. That can be very useful for post where they may want and have even finer control.

32 bit float would handle that, of course you and your mics would be cinders.

Maybe not the trusty 416's Howard

Depends on how you define “camera”.

Any iPhone 15 Pro (Max) is able to record up to four channels of IEEE 32-bit float as part of standard ProRES 422 at several qualities and frame rates - one example is Blackmagic Camara (since the first version).

The new Panasonic GH7 is also capable of recording IEEE 32-bit float internally (I assume a ProRES format is required), so…

Limited knowledge or outdated beliefs may hamper the argument.

Maybe you should focus on your own limited knowledge

That Panasonic, just released, requires a US$500 separate sound recording accessory to record at 32-bit float.

Apple’s iPhones are designed to make sound recording idiot proof. They include heavy compression, aka limiting of dynamic range. That’s why the waveforms of an iPhone recording of a New York City subway train, and of a MixPre recording of the same train, look entirely different. Yes, I’ve made comparative recordings. The iPhone significantly reduces the dynamic range. Use an iPhone to record the sound of a NY subway train and you’ll lose almost all of the low end, which is where much of the energy is, that a MixPre will record.

While I've only tried this comparison with an iPhone, I expect that all of the phone manufacturers are doing the same thing. I’m not sure that it’s even possible for an iPhone recording to exceed 0dB. It certainly isn’t common.

I’ve yet to see anybody claim that one can select 32-bit float recording on an iPhone 15 and take the resulting recording into a 32-bit float workflow. I’ll certainly be interested when and if somebody demonstrates this.

Depends on how you define “camera”.

Any iPhone 15 Pro (Max) is able to record up to four channels of IEEE 32-bit float as part of standard ProRES 422 at several qualities and frame rates - one example is Blackmagic Camara (since the first version).

The new Panasonic GH7 is also capable of recording IEEE 32-bit float internally (I assume a ProRES format is required), so…

Limited knowledge or outdated beliefs may hamper the argument.

Maybe you should focus on your own limited knowledge

That Panasonic, just released, requires a US$500 separate accessory to record at 32-bit float.

Apple’s iPhones are designed to make sound recording idiot proof. They include heavy compression, aka limiting of dynamic range. That’s why the waveforms of an iPhone recording of a New York City subway train, and of a MixPre recording of the same train, look entirely different. Yes, I’ve made comparative recordings. The iPhone significantly reduces the dynamic range. It’s why YouTube is full of phone recordings of NY subway trains, none of them clipping.

Use an iPhone to record the sound of a NY subway train and you’ll lose almost all of the low end that a MixPre will record.

While I only use iPhones, I expect that all of the phone manufacturers are doing the same thing. I’m not sure that it’s even possible for an iPhone recording to exceed 0dB. It certainly isn’t common.

I adore your style. Ahem!

I use(d) - since the first version - a Zoom F3 recorder as IEEE 32-bot float Audio Interface allowing the use of any of my Phantom powered /or not) external microphones to deliver beautifull sound.

It seems, that you now try to disown internall camera IEEE 32-bit float recording, just because an external interface is used. Why is that unacceptable?

Are you against ANY use of external audio interfaces on any camera? Inventing what could be construed as an artificial limitation. Will you also object to an USB microphone delivering 32-bit float directly to the iPhone USB interface, or do you even require analogue input converted to IEEE 32-bit float in camera in order to defend the view “no camera records IEEE 32-bit float internally”?

Just asking?

I’m more relaxed. If a standard 32-bit IEEE float signal (any number of channels) can be digested by the camera, and added into the recorded file in a standard, well defined way - I.e. ProRES up to four chann3ls - then that camera can record IEEE 32-bit float.

You’re free to have your own convictions, but please - try to moderate your condescending language. Even if you disagree!

I adore your style.

It seems, that you now try to disown internall camera IEEE 32-bit float recording, just because an external interface is used. Why is that unacceptable?

Got it, you take exception when your very own words are applied to you.

If you want to think that Panasonic’s US$500 charge for an external sound recording accessory is irrelevant to this discussion - in your world, apparently not even worth mentioning - you go right ahead

[quote="kfriis”]
I adore your style. Ahem!

It seems, that you now try to disown internall camera IEEE 32-bit float recording, just because an external interface is used. Why is that unacceptable?[/quote]

Got it, you take exception when your very own words are applied to you.

If you want to think that Panasonic’s US$500 charge for an external sound recording accessory is irrelevant to this discussion, you go right ahead [/quote]

How about the Zoom M4 microphone (whether the quality is acceptable or not)?

See: https://zoomcorp.com/en/de/mictrak-reco ... 4-mictrak/ (haven’t tested specs in real life, but Zoom usually is quite correct on the specs).

That “microphone/recorder” delivers - for instance - 32-bit float to a modern PC, Mac or iOS smartphone directly to the USB port. Do you - now - insist on 32-bit IEEE recorded from internal camera mounted microphones, or can you accept, that the world is changing?

How about the Zoom M4 microphone (whether the quality is acceptable or not)?

What about it?

The Zoom is just another 32-bit float recorder, although at US$340 street not cheap.

So is Diety’s PR-2 and various 2.4GHz wireless recorders from Røde, DJI, etc.

Some 32-bit float recorders will store the sound in a camera, some may not.

So what.

The first post in this thread is about a camera recording sound at 32-bit float and expressly differentiates that from using an external device.

If one wants to use an external device, but store the audio in a camera, there are already options.

If you want to do a show and tell exercise, go ahead

How about the Zoom M4 microphone (whether the quality is acceptable or not)?

What about it?

The Zoom is just another 32-bit float recorder, although at US$340 street not cheap.

So is Diety’s PR-2 and various 2.4GHz wireless recorders from Røde, DJI, etc.

Some 32-bit float recorders will store the sound in a camera, some may not.

So what.

The first post in this thread is about a camera recording sound at 32-bit float and expressly differentiates that from using an external device.

OK. Religious purity?

As far as I can see, the M4 allows direct 32-bit float recording to USB C; that’s typical for many Zoom products, even if they have an internal recorder. You don’t have to use it, you know, or is that against your principles?

Now, what about the Røde NT 5. Gen microphone.

It delivers - natively - a standard USB C iEEE 32-bit float signal (see https://rode.com/en/microphones/studio- ... u=NT1GEN5B) or do you object to delivery of the signal via USB on principle (even when designed by the microphone manufacturer).

Is electronic at source conversion objectionable on principle? How can you then defend an XLR based based analogue signal generated by Phantom 48V powered circuitry in the same microphone?

Do you require also a special form factor? Maybe dynamic only microphone (Phantom an abomination) for purity reasons?

As I mentioned earlier: The world is changing.

You may not like it. Even object to the changes ahead, but at least on iPhone 15 Pro you can record IEEE 32-bit float signals directly from a microphone (in the NT1 case a really good condenser with extremely low internal noise).

Now you need to come forward, and define precisely and in clear and accurate detail, what you regard as “true” internal IEEE 32-bit float recording in camera.

Or concede, that it is already possible to record IEEE 32-bit float internally in camera from a feed delivered by a single cable from a mass produced microphone of accepted quality.

This is one of those exchanges that puts me in a Bartleby mood

Now you need to come forward, and define precisely and in clear and accurate detail, what you regard as “true” internal IEEE 32-bit float recording in camera.

Or concede, that it is already possible to record IEEE 32-bit float internally in camera from a feed delivered by a single cable from a mass produced microphone of accepted quality.

Sorry Kurt, I’ve switched you off in favor of a convention in Chicago.

Looks like some good music tonight: Jason Isbell, Mickey Guyton, James Taylor.

ADDED
For people interested in music... The performances of Mickey Guyton and Jason Isbell are already on YouTube. James Taylor’s performance may be moved to later in the week.

If the paywall isn’t a problem, the NY Times published an interview with Jason Isbell today in which he makes interesting comments about the relationship between music and politics.

T

Sorry Kurt, I’ve switched you off in favor of a convention in Chicago.

Looks like some good music tonight: Jason Isbell, Mickey Guyton, James Taylor.

I get it. You’re an extremely busy man, so I’ll likewise make things extremely simple and clear for you:

1. Do you deny, that the LUMIX GH7 camera is capable of including IEEE 32-bit float audio internally into video recorded by the camera?

2. Do you deny, that the iPhpne 15 Pro (Max) is capable of including IEEE 32-bit float audio internally into video recorded by the device (example: Using the Blackmagic Camera App)?

A simple yes or no will suffice.

I definitely hope Blackmagic adds this. If we can record Raw video, why not have the ability to record 32 bit internally? Sure, most people record externally with recorders, but the way I see it, if we have the option to use XLR mics, we might as well have 32 bit as an option. And if people are worried about file size, they can always add a toggle to record in 24 bit instead for those who don't want it.

I definitely hope Blackmagic adds this. If we can record Raw video, why not have the ability to record 32 bit internally? Sure, most people record externally with recorders, but the way I see it, if we have the option to use XLR mics, we might as well have 32 bit as an option. And if people are worried about file size, they can always add a toggle to record in 24 bit instead for those who don't want it.

For gear targeted at the run-and-gun crowd, that will probably be a game changer on the audio front. In run-and-gun anything compact that needs not being placed onto intricate or even flimsy rigging is a bonus.

The handlebar on the LUMIX GH7 is not ideal in my personal view. Two built-in mini-XLR in the critter AND audio-in via USB-C would be a “narjs” and more compact solution. Maybe Blackmagic is aiming at higher market segments now.

Regards

This video just dropped today all about 32-Bit Float.


Sent from my iPhone using Tapatalk

This video just dropped today all about 32-Bit Float.

Sent from my iPhone using Tapatalk

Barry Manilow: Could this be the magic at last? Could it be magic?
Curtis Judd: No.

This video just dropped today all about 32-Bit Float.

Sent from my iPhone using Tapatalk

Barry Manilow: Could this be the magic at last? Could it be magic?
Curtis Judd: No.

You’re misquoting Curtis Judd - or at least “adjusting” his messages to an agenda.

The short form is: If you’re working in controlled environments - like he usually is - the need for using IEEE 32-bit float over 24 bit LPCM is not prevalent.

Now, one person run- and-gun, some situations (rallies, processions and parades or open air performances come to mind) are not what I would call “controlled environments”, Which Curtis Judd ALSO recognized as a very “different use case” from his well controlled YouTube studio.

The problem is, that you seldom - as in never ever as a mere mortal - get a chance to ask the President of a country to stop, and please start the speech over again. Or try to ask a Carnival in Rio procession to “back off 200 meters/yards and start over” just because, their enthusiasm overwhelmed the 24-bit limit you had set (and input clipped, but not the microphones)

IEEE 32-bit float doesn’t solve every bad situation, but if the microphone capsule is not screamed out of wack and front end electronics allowing (in my case up to +24dBu which will do in most cases and most microphones , a bit of latin enthusiasm can easily be recovered in post. If 24 bit LPCM was used, that would be impossible without riding the input levels - often impossible if you have to handle the video end too in quickly varying lights, distances and other “lively” aspects.

It’s like a travel insurance. You may not need it, but it’s sure as h*ll nice to have, if a local bus rams you in a foreign destination, and you survive

The “office boys” of video (most studio based YouTubers come to mind) can even do retakes to their hearts delight, if something was a bit off (settings or otherwise).

Regards

He seems a bit biased.

Never mentions that you cannot only recover 'clipped' signals but also raise very low signals without raising noise floor as well, as happens with quiet 24bit recordings.

He claims 32bit float can still clip. Then goes on to explain that when the microphone clips, so does the signal. Which is true for any recording, 16, 24 or 32bit float, but somehow he presents it as a fail.

He claims the microphone's dynamic range limits 32bit float anyways. I guess it does if you want to record an actor eating a cheese sandwich and a fighter jet taking off at the same time.

He claims 32bit float recordings only work with top notch preamps but it works with mediocre preamps as well, especially for content creators who do not have a dedicated sound person.


All in all I think this guy misunderstands that 32bit float recording is not about better audio (which btw it is for amateurs or creators, as signals will always be perfect) but a completely different workflow on set, freeing capacities while eliminating a point of failure.

He seems a bit biased.

Never mentions that you cannot only recover 'clipped' signals but also raise very low signals without raising noise floor as well, as happens with quiet 24bit recordings.

He claims 32bit float can still clip. Then goes on to explain that when the microphone clips, so does the signal. Which is true for any recording, 16, 24 or 32bit float, but somehow he presents it as a fail.

He claims the microphone's dynamic range limits 32bit float anyways. I guess it does if you want to record an actor eating a cheese sandwich and a fighter jet taking off at the same time.

He claims 32bit float recordings only work with top notch preamps but it works with mediocre preamps as well, especially for content creators who do not have a dedicated sound person.


All in all I think this guy misunderstands that 32bit float recording is not about better audio (which btw it is for amateurs or creators, as signals will always be perfect) but a completely different workflow on set, freeing capacities while eliminating a point of failure.

NO!

Curtis Judd really knows, what he is talking about, but most people simply do not listen to WHAT he says and the QUALIFICATIONS!!

Look at the video again.

Curtis stresses - again and again - the precondition “FOR SPOKEN WORDS”. And he’s completely right in that scenario (including handing material off to others).

In ANY PROFESSIONAL ENVIRONMENT, where you deliver material to “upstream processing/editing” you ALWAYS get the COMPLETE SPEC SHEET for delivery format. AND DELIVER THAT! Period.

If part of the task is delivering “original recordings”, that may pose problems, if you use 32-bit float. To use a completely different example: You probably can’t deliver any tire size and type to a Ferrari workshop!

The man is correct. We can differ on minor details, where taste and preferences may come to play, but the gist of his message is correct.

FOR SPOKEN WORD!

No buts!

There are other use cases. I mentioned a very different one.

That does not make Curtis Judds presentation bad at all, if people just cared to LISTEN, instead of seeking the “holy grail” in audio. ANY choice is a compromise based on actual use case requirement, “guru’s” and wishful thinking be damned

It’s about choosing the right tool for the task at hand. ALWAYS!

Regards

I mean I disagree.

He doesn't mention a whole range of advantages (spoken word or not). First and foremost, the workflow, but also that you can raise low levels without raising noise - super important for spoken word.
He misrepresents the importance of dynamic range of microphones (which he should perfectly know is irrelevant). And he claims without top notch preamps 32bit float recorders aren't working beneficial, which is simply not true, spoken word or not.

And at the end he asks to buy his courses.

I mean I disagree.

He doesn't mention a whole range of advantages (spoken word or not). First and foremost, the workflow, but also that you can raise low levels without raising noise - super important for spoken word.
He misrepresents the importance of dynamic range of microphones (which he should perfectly know is irrelevant). And he claims without top notch preamps 32bit float recorders aren't working beneficial, which is simply not true, spoken word or not.

And at the end he asks to buy his courses.

Let me add a simple example (I had to perform a few tests some days ago):

Including Røde Wireless Pro (TX only), Zoom F2-BT and DJI Mic 2 (TX module only).

ONLY interest is the recording side. I use the receivers to feed my headphones (monitoring, what may haven 20 feet away). Not recording severely restricted 24-bit receiver output (Curtis is correct about that).

I will not bother you with a lot of details (I don’t test for free;-), but a few key values are interesting. All tests simultaneously, same material (sweep, burst, tones) in near same position - centered around one speaker.

Max analogue input on TRS:

Zoom F3 and Zoom F6: +24dBu
Zoom F2-BT: -5.5 dBu (a roughly 30 times lower signal)
Røde Wireless Pro TX: No factory specs available (have asked). Max level around 2-3dBu higher than F2-BT
Røde Wireless Pro II TX: 0.1dBV (according to Røde support). Input easily clipping.
DJI Mic 2: -6 dBV

Lavalier (TRS input) Røde Wireless Pro TX

Røde Lavalier II: None of the TX modules were overloaded (but much noisier than “alternative”). Includes dead cat.
Noise: -81dB
Maximum level: -31.5 dB
Dynamic range: ~49dB

My “alternative” Lavalier: Slight clipping (or capsule overload) on DJI Mic 2, and easily overloading the Røde Wireless Pro TX. Inc ludes dead cat.
Noise: -83dB
Maximum level: -30,7 dB
Dynamic range: ~52dB

Special test case oriented toward a “live session”.

Very different environment and NOT comparable with below more controlled tests:

Dynamic range measurement (32-bit float), DJI Mic 2 TRS:

DJI Mic 2 TX built in microphone with dead cat (RMS):
Noise: -85.6 dB
Maximum level: +12.5 dB
Dynamic range: ~98dB

DJI Mic 2 TX “preferred lavalier via TRS”:
Noise: -75,1 dB
Maximum level: +13.9 dB (slight clipping or capsule problem?)
Dynamic range: 89dB

For comparison only (official Røde data, only controlled test environment values):

Røde M5 condenser:
Noise: 19 dBA SPL
Maximum SPL: 140 dB SPL
Maximum output: 13.5dBu
Dynamic Range: 121 dB

Røde NT1 condenser:
Noise: 4.5 dBA
Maximum SPL: 132 dB SPL
Maximum output: 8 dBu
Dynamic Range: 127.5 dB

No wishfull thinking required.

Just actual tests and measurements in ordinary environments.

Regards

What are you trying to say?

What are you trying to say?

Nothing. You can read, can't you?

It's not religion or rocket science.

Read and conclude yourself.

I'm adding a set of dynamic range measurements involving the Røde Wireless Pro TX modules (32-bit float)

Just for completeness.

Røde Wireless Pro TX modules (32 bit float):

Røde Wireless Pro TX Lavalier II with dead cat (RMS):
Noise: -72.3 dB
Maximum level: +16.7 dB
Dynamic range: ~89 dB

Røde Wireless Pro TX “preferred lavalier via TRS”:
Noise: -74.8 dB
Maximum level: +15.3 dB
Dynamic range: ~90 dB

Completely different environment, days earlier, so direct comparison is NOT possible. As I alluded to earlier, I'm not born into this world to deliver actual real life measurements.

NOT a scientifically reproducible environment treated for technical reference measurements.

Just quoted for completeness.

Regards

DR of microphones has little to do with the true benefits of 32bit float recording (unless you have really crappy ones, but that goes for 16 and 24 bit as well). If you don't place your mics like an idiot, it does not matter.

I don't know how often it has to be repeated:

It is not about better audio quality, it is about not having to control gain during recording anymore.

The benefits for on-set field recording should be clear to anybody who does on-set field recording, the primary use case of 32bit float recording (these are all field-recorders).

I don't know what Curtis Judd's actual production credits are, but if he's too skilled or too well prepared and crewed to need 32 bit float, hallelujah for him.

The "use cases" for 32 bit float have been noted many times in the course of this thread, and they don't include professional sound crews working on controlled sets. For parties without the ability or leisure to constantly monitor and adjust levels, 32-bit will at least ensure that, barring the extraordinary, the recording won't clip and will never drown in the noise floor. That's going to be worth something to somebody. To a solo operator, it could be worth a whole lot.

Curtis reports that he's not "100% sure" that there are no "other costs to 32-bit float", but until he is and makes his announcement.....

DR of microphones has nothing to do with the true benefits of 32bit float recording.

I don't know how often it has to be repeated:

It is not about better audio quality, it is about not having to control gain during recording anymore.

The benefits for on-set field recording should be clear to anybody who does on-set field recording, the primary use case of 32bit float recording (these are all field-recorders).

Dynamic range per se, has nothing to do with 32-bit float performance and abilities.

Real life dynamic range certainly has. And that’s placed BEFORE any 32-bit float circuitry of whatever design.

Good interfaces have a noise level of down to around -127 dB or even less (overload at 4.4 dBu) other settings deliver -107 dB interface noise and 24.4 dBu overload level. A maximum dynamic range on the audio interface of 131.4 dB. Whatever you decide for “overload level”.

NOTHING alters that. iEEE 32-bit float or not!

Then comes microphones.

The NT1 has a minimum noise of 4.5 dBA SPL, extremely low. An extremely high output level for “only” 132 dB SPL (max dynamic range 127.5). Should be within the interface capabilities, but the interface will overload “normally” +4.4dBu settings (the microphone delivers 8 dBu max at very low distortion). If more noise is allowed, then input overload will not be a problem, but then you will restrict dynamic range to around 107 + 8 dB in the front end - BEFORE any conversion to 32-bit float.

32-bit float ONLY can handle “practically unlimited” signals AFTER the interface input stage. A stage, that in many instances has limited to non existing gain adjustment.

Not clipping is far more valuable in many situations, than very low noise, but still…

If a hot microphone overloads your interface input, 32-bit float will not prevent that.

If a robust microphone, like the relatively cheap M5, can handle 140 dB SPL, and easily overload the mentioned standard input settings pf 4.4 dBu, when delivering 13.3 dBu without getting any sweat, a dynamic range of 121dB certainly has some influence (and forcing you to accept higher noise levels from the interface if max +24.4 dBu is selected in order to avoid any clipping).

Now, sound levels usually don’t stay at high SPL’s for longer times, where humans are around - but peak is not unheard of - so there is no guarantee, that the higher interface noise will be completely negligible, if you’re forced to lift very silent passages in post. And your maximum dynamic range is around - hmmm - (wrong 101) 107+13.3 dB or 120 dB give or take - to avoid clipping in this example. If my calculations are correct (in second attempt)

The real benefit is, that if your setup is conservative enough (maximum input level set to +24.4 dBu not +4.4 dBu in this real life example) the M5 microphone capsule will go bonkers before the interface at more than 140 dB SPL. At the cost of higher noise, at near silence.

IF you can’t concentrate on riding the levels - as in most run and gun situations - that”s the best guarantee you have of getting good sound, whatever level presents itself to the gear in real life. Really valuable in uncontrollable environments. iEEE gives access to this relative freedom, but that’s all.

The lavalier examples I used were purely aimed at illustrating typical talking head best case situation with normal enthusiast gear in real life. Clearly illustrating Curtis Judds arguments “FOR SPOKEN WORD”. All the interfaces used - and only front end TX IEEE 32-bit float recordings and devices involved.

But everything you write about mics applies to 16 and 24bit recording as well.

But somehow it's now used as an argument against 32bit float recording when it was always, well, a mic(ing) issue with 16/24bit.

But everything you write about mics applies to 16 and 24bit recording as well.

But somehow it's now used as an argument against 32bit float recording when it was always, well, a mic(ing) issue with 16/24bit.

Yes! Front end capabilities (noise compared to max input level) and microphone specs (noise and max input translated to max juice out) is really all, that counts and limits your options (except for 16-bit).

The ONLY difference is, that MOST better enthusiast interfaces have a very low noise floor, and can be set to accept very high input levels - seldom the full range without some kind of adjustment/attenuation at play.

But within that range, you do not have to control levels.

16-bit is especially problematic; if you aim to create a reasonably conservative headroom, noise becomes a problem at low levels. And you still have to keep an eye on real life levels.

In 24-bit you can be more conservartive relative allowed headroom, but still - zero dB is zero dB - not +4.4dBu or +24.4 dBu as in my example.

If your interface can handle the levels (M5 13.3 dBu is in the extreme end) and not get into noisy behaviour at very low levels, you can adjust when using 24-bit (but the problem will be the same as for 32-bit, if you have to deliver unadulterated original material to a third party).

The more "economical" 24-bit interfaces, are typically not in the same league, as good 32-bit float interfaces. They don't have to be, since you CAN "ride the input levels" and fine tune input to fit realities.

Given a VERY good frontend in IEEE 32-bit float gear, you can be excused from level duty. In a few cases, you have to decide (based on microphone capability and how "hot" it is), whether to aim for extremely loud sound levels (and some minor noise) or very high sound levels with low noise. With many "hot" microphones - Røde NT1 and some Lewitts etc - are prime exceptions - you often have a far higher relative noise level, than the interface can deliver). That's more easy in real life.

It's as I stated before: IEEE 32-bit float is similar to a travel insurance. You may not need it at all, but if real life turns nasty, you're glad you had it.

I'm ONLY talking single person "run-and-gun" situations, not huge cinema crews in the hundreds with several trucks from Pinewood studio and c catering facilities. Or super controlled (YouTube) studio environments, where you can - if a bulb pops or something alike - redo (part of) a take, if needed.

Regards

Barry Manilow: Could this be the magic at last? Could it be magic?
Curtis Judd: No.

You’re misquoting Curtis Judd - or at least “adjusting” his messages to an agenda.

No, I wasn't. Judd states very clearly in that video that it's not magic and won't give us world peace. Apparently, I (and Curtis) have a sense of humor and you do not.

You can buy a recorder in the amount of time it would take Blackmagic to add that. And I don't want the added cost.

... Judd states very clearly in that video that it's not magic and won't give us world peace. Apparently, I (and Curtis) have a sense of humor and you do not.

Precisely. Judd knew that a lot of his viewers don’t want to hear what he says in that video, and he kept it as light as possible. That comes through in his presentation.

His video follows Allen Williams’s (YT channel Sound Speeds) detailed criticism, two weeks ago, of Røde/DJI/Etc Bluetooth wireless, which are now being marketed with “32 bit float”. Courtesy of saturation marketing and “influencers”, those units are almost certainly the main way that people are acquiring 32 bit float.

Both videos are long overdue.

Sound Pro Reacts to Brianized "Don't Use Wireless Mics" Video

You can buy a recorder in the amount of time it would take Blackmagic to add that. And I don't want the added cost.

I’ll buy a different brand of camera before I’ll pay for this “feature”. On the rare occasions that I use 32 bit float, one of which I described in detail earlier in this thread, I have a perfectly good MixPre that I have confidence in.

Funny! Nobody here ever said it was "magic". Neither does Zoom nor Sound Devices. Did that claim come to Curtis Judd from youtube ether? If anything, the posts in support here are on the timid side. Arguing the obvious over and over again may not merit quite so much restraint.

I’ll buy a different brand of camera before I’ll pay for this “feature”

That'll stop 32-bit float development for sure!

Tascam is about to release a small 2 channel 32-bit float recorder with TC gen, sync BT remote monitoring and USB-C audio interface etc.

https://www.instagram.com/tascam_offici ... imEo5qljb/
https://www.instagram.com/tascam_offici ... 0no0Ii62j/

Something big is coming for videographers, field recordists, and content creators...

• 2 XLR/TRS Combo jacks
• 32-bit float recording up to 192kHz
• Compact sturdy body that fits in videography rig
• USB-C Audio Interface
• Generate and sync/receive time code
• Jam-sync ready for other time code devices
• Accommodates wireless headphones
• Ability to Control/Monitor up to 5 units simultaneously

I think it is time for a real case, showing what I'm talking about.

The other day I was outside the Deoksugung Palace in the center of Seoul at the changing of the guard.

I used the DJI Osmo Pocket 3 (I also have the two previous versions) and an associated, linked DJI Mic 2 with dead cat (placed inside a chest pocket with "flap" protecting against the rain. The setup controls the microphone/recorder too (32-bit Float). When recording starts, the mic does, and when recording stops, the mic does too. That works amicably well. There is of course sync in post, but an extra "IEEE 32-bit to sane usage", like -23 LUFS EBU 128 etc without clipping peaks is required - no problem).

Here statistics from the recording:

Code: Select all

===========================================================================
Statistics Report                                  Sat Sep 14 21:10:13 2024
---------------------------------------------------------------------------
/Volumes/990Work4TB/KoreaMedia/DJI/20240913/Mic2/DJI_02_20240913_105034.WAV
---------------------------------------------------------------------------
Number of Channels                      1 channel
Sample Rate                         48000 Hz
Resolution                             32 bits
Audio Duration                  03:20.075
===========================================================================
Statistics                         Ch.  0         
---------------------------------------------------------------------------
Minimum Sample Value               -47330         
Maximum Sample Value               +35018         
Sample Peak                         +3.19 dBFS   
Possible Clipped Samples              222         
DC Offset                           +0.25 %       
---------------------------------------------------------------------------
True Peak                           +3.20 dBTP   
---------------------------------------------------------------------------
Minimum RMS Power                  -45.14 dB     
Maximum RMS Power                   -3.33 dB     
Average RMS Power                  -34.82 dB     
Total RMS Power                    -30.99 dB     
---------------------------------------------------------------------------
Integrated Loudness                -37.37 LUFS   
Short Term Maximum Loudness        -22.86 LUFS   
Loudness Range                      +9.60 LU     
---------------------------------------------------------------------------
N                                 9603600 samples
===========================================================================
RMS Settings
---------------------------------------------------------------------------
RMS Window Size                        50 ms
Wave Type                       Sine Wave
Account to DC                         Yes
===========================================================================


and an "image" of the recording in question:

DJI Mic 2 w dead cat recording of Changing of the Guards (first part with drums)

Skærmbillede 2024-09-14 kl. 21.09.20.png (362.79 KiB) Viewed 1785 times

Note, that minimum RMS is not "technical noise", but the average minimum "noise" in form of people and friday inner city traffic noise at around 11:00 before noon in a mega city like Seoul, just prior to the Chuseok celebrations. There's heavy, everyday traffic going on just a few meters behind my position (as far in the front, the front, that security allowed).

There is no clipping here ( I have just not recovered the peaks yet), but there was no way in hell, I could be prepared for these conditions (as a normal visitor or tourist). I could revisit, but still, if using an ordinary 24-bit recorder, I would be inclined to select around -40 dB "average" recording level in order to secure no clipping in the drum sequence.

Your gear is of course good enough, and you would have to go through the same motions, as I had to, when "massaging" source into usable "original" audio material. This is not the question. The question is, would you have done it?

In this case, I had the best of both worlds. The simplicity of IEEE 32-bit float (during recording) and the reasonably safe assumption, that my recording would be well and safe at all levels (within the around 100 dB dynamic range the DJI Mic 2 microphone/recorder can handle with previously verified certainty - maybe more in reality).

This is not theory. It's a moment in time during travel in South Korea, where action was frozen in video and sound for posterity. Whether quality is good enough for the purpose, only I can judge.

No video, yet. Uploading via the hotel WiFi is of no practical use for large files. It takes around 1 minute per second of video material at h265 75 mbps 4k 25 fps 2020 HLG 4:2:2 24-bit LPCM stereo (had to start test at a known state). Let's say, uploads are extremely slow and choppy to boot.

Regards

… There is of course sync in post, but an extra "IEEE 32-bit to sane usage", like -23 LUFS EBU 128 etc without clipping peaks is required - no problem)…

Thanks for the details.

That audio waveform tells the story. 32bit float, within the limits of your mics, ensures one is prepared for the unexpected peaks that may occur. I also always apply -23 LUFS EBU 128 adjustments in post.

Thanks for the info you two!

I had performed a simple google search on stores in Vancouver and if I had seen Trew, I would have looked them up because I've been to their store in Los Angeles. Now that I know they're in Vancouver, I'll see if I can get a ride up with my friend and go pick those up. It would be nice to get the International version.

To close my anecdote. I did go up to Vancouver to Trew Audio. And they were sold out of PR-2s because apparently, I'm not the only person in the PNW who had the idea to get involved in the audio smuggling business. Hilarious!

Darn! Hopefully they’ll set some aside for you when they have stock again.

Darn! Hopefully they’ll set some aside for you when they have stock again.

I have an interview to record next week so I just bought two more of the US ones. It'll be fine. I did enjoy some amazing sushi while I was up there.

[MAPLE LEAF]

I do love the PR-2. It's great. My only complaint comes to not being able to monitor audio while it's recording.

Now back to in camera 32-Bit Float Recording. Shot with the P6KPro the other day, and still wished I had it. Ran the Sennheiser 416 boomed and the Sennheiser G4 Wireless Lav into camera. Mainly talking head so set gain and it didn't change much. But every once in a while I would need to gain it down for loud moments. Would be so nice not to have to ride gain levels on camera.

I'm pretty sure it will be in cameras one day and it will probably be low tier cameras that sport it first as it's a very beneficial feature for content creators.

With that out of the way, 32bit float recording may be more expensive and difficult to implement in a camera.

I don't know..

It already comes in tiny cheap(ish) recorders these days.

It still takes two ADC instead of one, so a little more cost, space and knowledge. I'm sure it will come sooner than later. maybe even in a prosumer / creator cam first.

I do love the PR-2. It's great. My only complaint comes to not being able to monitor audio while it's recording.

Now back to in camera 32-Bit Float Recording. Shot with the P6KPro the other day, and still wished I had it. Ran the Sennheiser 416 boomed and the Sennheiser G4 Wireless Lav into camera. Mainly talking head so set gain and it didn't change much. But every once in a while I would need to gain it down for loud moments. Would be so nice not to have to ride gain levels on camera.

I'm going to be running PR-2s (recording satanic 32-bit float) and running a boom mic back to the camera as a backup track so I can get TC. I have noticed that with the BMCC6K that it records pretty low and I have to boost in Fairlight, but the sound is good both there and when I've imported audio files into Audacity. I'll be watching the levels, but am not too concerned.