You are a pro, and it is second nature for you to set the gain in a recorder. For me -an amateur- setting the gain was my worst anxiety in the entire filming process. Every single time I had a sound problem, it was always because of wrong gain setting. 32-bit float is not marketed for pros. You guys can record amazing audio with 24-bit. For non-professionals, however, 32-bit float IS a game changer because it offers us peace of mind knowing that we will not clip. Great video. As always. Respect.
Yep, this is exactly what ZOOM and Sound Devices like to hear. On a separate note, I think of setting gain as very much like setting exposure on a camera. Definitely a skill worth developing (and keeping 32-bit float as a safety net). Another thing many people do is push their gain way too high given the quality preamplifiers we have to work with today. Peaking around -20 is probably what most of us should be aiming for so we'd have far fewer issues with clipping.
I don't get this, because it seems to me that if you can't figure out how to set gain then you also don't know where you are within your microphone's dynamic range, and _that_ makes it almost impossible to get mic selection and placement right. It seems to me that understanding where your levels are is a prerequisite to any sort of high-quality audio work, not a nice-to-have that's only relevant to the final recording process. After all, you have to avoid saturating your _mic_. And that's where the marketing for the likes of Zoom really misses the boat. All of their professional-quality sample recordings were undoubtedly created by people who understood exactly where they were within the mic's range and therefore its output voltage.
Not clipping is only one small part of getting good audio, that's only talking about a rookie mistake. There are lots of big mistakes one can make, and that's just the most obvious one that's usually very easily noticed and consequently solved. Taking this one issue off the table doesn't get rid of a host of other potential problems. By far, the biggest mistake -amateurs- people make is poor microphone placement, including booms, plants, and lavs.
2 หลายเดือนก่อน +5
In my area i often have choirs or small music groups that recorded themself, did not level the audio at all or too quiet or too loud, and then i had to mix and repair clipd and super noisy signals ... or i recorded concerts where i gave recorders to the sound guys, just to have a stereo direct feed of the bands on stage, we leveled everything together but later in the evening everything got louder and clipped everything, or at weddings sometimes i have to be super fast and have no time to level again, including going from people talking without a microphone the the audience cheering ... back in the day this ruined so much, but since ichanged to 32bit on everything, this is gone forever. That is why i love it.
Doing weddings all by yourself will a lot of the times not allow you to check if you're ok with audio levels, and i did a couple of times had the audio levels too high because a priest decided to go way up with his voice mid service and i didn;t notice because i was too busy filming with the cameras. For me, to be able to recover spoken word from some instances, be they as rare as they are, is actually very useful.
Yup, using the right tools for the right circumstances. Nothing wrong with that. Otherwise, people posting videos on TH-cam really should first study proper film and cinematography and not take the lazy shortcut of digital videography. Celluloid transfer rules any day over fancy digital schmigital sorcery!
I found 32bit super useful for thunderstorm recordings. Especially when I left my droprigs out in the nature for a several days and I can't adjust gain. It's just perfect for this because I can capture a super quiet atmosphere and when the lightning strikes very close - I just pull that part down in postpro to balance everything without any distortion. I've also found it useful when the wind hits the mics a bit harder. In post-production you just high pass that part and rest of the frequency spectrum will survive without any damage. It's life saver sometimes. But otherwise, for normal ambiences where I not expecting sudden loud noises - 24 bit is more than enough. Btw great explanation as always Curtis!
Important video, finally! I've been emphasizing the clipping potential of most lavs for years, despite the simplified talking point that 32-bit float insures against clipping. Also, it's so important to debunk the oversimplification that a mic will clip "only next to a jet engine." Happens all the time. One last thing to suggest is that tools like De-clip in editors like iZotope RX sometimes repair clipping reasonably well. Also: should have recommended the completely separate tier of lav microphones from the likes of Countryman and DPA that are designed for loud volumes (e.g., singing and sports) sacrificing dynamic range on the quiet end.
This may not be the strongest argument. A mic will clip when overdriven regardless of bitrate. At best, the argument can be that SOME people MAY incorrectly think that their mic may not clip under 32-bit.
Thank God you weighed in on this. I did this same topic and got so much unbelievable push back. I don't understand why people PREFER to spend money than just learn how to use their equipment properly
Thanks, John. I'm getting some push back as well. I also suspect that there are some nuanced costs to most 32-bit float implementations but need to do more digging. Keep making great education over there, friend!
I remember that video. I thought It was one of the best practical presentations on how good the 24-bit tech already is. But I will also say that the answer is already in your question: paying takes less effort than learning, so that's why.
Appreciate the honesty and in-depth description. I was definitely sold that 32-bit was magic and was the way out of not having to worry about setting levels and fiddling. I'm thankful for your reality check - you'be probably saved me hundreds of hours of future headaches by teaching this.
Recording spoken word in fairly reasonable environments with a MixPre3 II - I use 24bit and do gain staging. Yesterday, as an example, recording "spoken word dialog" of a riverboat ferry captain, on a very windy day + a loud noisy boat I used 32bit float. I still did as much gain staging I could in that unpredictable place as I could and used the fader when needed. Listened to the audio last night and the spoken word came through nice and clear. Nothing of significance to address in post. I probably could have used 24bit, but using 32bit in that situation I at least had the feeling that the margin of error was going to be in my favor.
Still, the most significant factor is that you started with a recorder that has terrific components throughout- great preamps, lots of headroom on the busses, negligible noise and distortion, good metering... ..plus, you rode the faders intelligently.
Excellent video- thank you, Curtis. People overlook that low end 32-bit recorders (ie; Zoom H-series) have nasty built-in mics and preamps. Plus, handling noise must be avoided- it's way worse than any potential benefit from 32-bit float. 32-bit files can be useful, but only if the client knows what to do with them, and we nail all the other factors that create a good recording.
Thanks. I have a confession to make: I record almost always in 24-bit. I am pretty decent at setting gain so I'd rather not do extra work in post production with a 32-bit float recorded file that exceeds 0dB. But the main thing you'd have to do for 32-bit float recordings is find those cases where the audio goes over 0dB and either compress them, use clip gain to lower them, or drop the levels of the entire recording so that those bits are below 0dB.
I have two quick thoughts: 1. No matter how well you set your gain on any bitrate, even 16 or 24, your microphones can still receive too much pressure. So that alone is not a slam-dunk argument against 32-bit float. 2. For field recordists who set up long-term unattended rigs, they very frequently talk about worrying about an unexpected event/noise clipping their recording. None of these technologies are wonder drugs. You use the best tools you are able to use for the right purpose, and you learn their ins and outs to apply them properly.
When you are working on your own 32bit float is a godsend. Because recording levels don't have to worried about, you can concentrate on all the other aspects of film making including, microphone placement and all the other thing you mention. I look forward to the time when cameras have 32bit float recording built in.
This was lots and lots of fun. Crystal-clear, wide-spectrum, HDR coverage (32-bit?) of a neat topic. I enjoyed it from the first frame to the last. Thanks! This is VERY good to know!
Thanks for making this video as I learned a lot from your videos as a sound recordist and learning lighting to make me a better boom op! Love your work
I think you're spot on when talking about recording spoken word and dialog, particularly when working indoors. But we don't always work in doors. And as a recordist, you often do not know when someone might start shouting rather than talking in a normal tone of voice. If it's sudden and you don't know it's coming, it is a game changer in that you now still have workable audio. With a typical 24 bit recording, even if you had the gains set perfectly lets say at -18db FS, it probably clipped at least a little bit. I do agree it's not going to solve all the problems. But it solves a big one if you're working fast and as you said, are using good equipment. Another point I think you kind of missed is what happens when you have recorded your audio too low or there is a really quiet passage. You sort of touched on it, when talking about quantization errors, but probably should have dived a bit deeper. Having 32 bit float not only means you can bring down audio that went over 0db FS, but it also means you have a lot more bits of resolution at low levels, so if you recorded too softly, then you can boost the audio up without boosting the self noise as much. My point is, we are not all recording someone or even a few people talking. Some of us record music concerts or ambient sound or all sorts of things that have far more dynamic range than what can be captured with a basic 24 bit recording. And yes it is more than what will be reproduced by the end playback device. But at least we have a better chance of fixing it in post than we did before. I respect your opinion quite a bit, but for certain situations with good gear 32 bit float can be a game changer. It has allowed me to save a lot of recordings that I only had one shot at, that otherwise would have been worthless or very problematic to fix. Some simple volume changes is still a lot faster to fix in post than noise reduction passes and anti clipping passes and all the other tricks we used to do.
One note: I’m not sure that 32-bit float results in lower noise recordings. My experience is that there’s still plenty of ambient noise and self noise depending on the mic and recorder.
A quite educational video, on the aspects of recording technique that you still need to get right in order to benefit from a technology like 32-float recording.
Thank you for that great explaination! As a consumer recording voice only i just use a used Zoom H1 and set gain lower than -18db. So it never peaks and as i don‘t hear the difference i live with some quantization errors. I don‘t record classical music but voice only, and as long as i am in a quiet room and i don‘t get noise from clothes everything is fine. And all much better than the microphone in the camera or fideling with wireless microphones that may work or may not work. Mixing video and sound from external records is so easy this days with davinci resolve….
Thanks. That is very useful. As a solo-creator / beginner often working in uncontrolled environments, I've gone for 32bit to avoid gain issues. I have a Zoom F2, DJI Mic 2 and a Sennheiser MKE 600 (which I plug into the DJI Mic 2). The problem, however, is low volume for spoken word and there is no way of adjusting this on my equipment. Gain can only be raised in post, where it runs into issues. For example, pushing the gain high enough to make the spoken voice clearly audible often leads to the recording peaking, even though it doesn't sound like it's distorting. It would be great to get a post-production workflow for addressing low volume spoken word on 32bit.
Curtis, I'm curious what your limiter settings are on your 888 for spoken-word/dialogue. Threshold, knee, attack, release, ratio? Bus vs. channel? If you could explain what settings you use and why you chose them, that would be really helpful. I think I've got a decent handle on it, but would love to expand my understanding if possible. Thanks!
Thanks so much for this fantastic explanation. What a shame though that it'll likely struggle to get the reach of some sponsored influencers sharing rhe marketing hype. Your knowledge is very much appreciated!
Thanks for this very clear explanation! You constantly refer to "Spoken Word Audio" - what is the typical dynamic range for live musical performances? Thanks again!
You are the best. It could be great to see a review on Zoom M3! nothing much about it really. It's the cheapest 32-bit setup I've seen. I really want to see you a review from you on that product. Thanks a lot
I've done a lot of recording of individuals speaking in front of a live audience. As only one microphone is used, I would only need to record to one channel. But I always use two and set the output level from the mixer for one of the channels a good bit lower. The higher level channel is set for the "optimum" level and in case of overload, I can replace that short segment using the other channel. I've seen some recorders advertise a feature to automatically record a lower level signal for the same purpose, but I think it's mostly low end equipment that offers that, so I've never tried it. Use your ears, not your eyes. If the waveform display in your DAW or editor looks clipped, listen to it to determine if it's actually objectionable. As Curtis was saying, the recorder records the signal it's fed, so if it's already distorted, well it is what is, "it" being you failed to use a better mic or you failed to train the speaker or you didn't do whatever it was that could have prevented the distortion.
I've seen that safety track feature, it wasn't on what I'd call low end, it was quite a bit more expensive than entry level pro gear. But it was specifically for one-man-band type setups. It's like those 'composer' synths, they cost 5 times or more what a 'normal good' synth does, but they don't do the same thing. A 'composer synth' you can play a big band setup on by yourself, it's not an instrument, it's a band simulator. It's a drum machine, base synth with sequencer, a synth and multiple lead instruments, all sequencable and live playable at the same time. Probably need a LOT of time to learn how the thing works, or take classes or something. Cool things - sorta, but the multi thousand price tag etc plus steep learning curve is something stopping most from getting into them I think. Except obviously the people employed as musicians using them for 'live' music.
9:28 I see the difference of having 32bit vs not having it as (all other things being within spec): "you WILL still have to do the work in post," but at least you CAN still do the work in post.
6:54 The reason we use floating point format is because it allows us to sacrifice precision for range and vice versa - the lower the integral part value, the bigger the fractional part's precision is. When using an integer format the difference between values in quieter parts become relatively smaller and the quantification errors are getting more significant, as you mentioned. Float jest let's us utilise those 4 bytes effectively for sound recording application. But as you said - that's only 1 part of the equation, the hardware needs to be adequate.
As a run a gun team who shoots cars for a living 32 bit recording in the DJI MIC2 is a godsend. We don’t have a dedicated sound guy so I have to set the old recorder (zoom h6) to a very low level in anticipation for the loud engine / exhaust.
@@curtisjudd I pair them with the Rode Micro and its pretty usable. It does lose out a bit of low end, but I haven't found anything as small & easy to set up with a decent SPL. Side note, I noticed the 2nd generation of the mic has a far lower Max SPL
Great video Curtis once again. I have a question please. Pro photographers with super wide exposure settings still like to carry a light meter. In pro audio, do you ever use a DB meter when recording louder content?
Hi as usual amazing video wonderful information. may I ask you to make a video if you didn't already about how do you convert 32 bit float audio to 24 bit float in the most common software so that everyone can know how to do such a task as you mentioned that not all post production houses want to get 32 bit float sound so in case we need to convert it what is the best and fastest way to do it. and with what software.
I wonder when will they make an 32 bit lav mic capsule, the zoom m1 has a 32bit float shotgun mic but I've only seen like 2 videos that reviewed that mic , so it must not be as great
Mic capsules are analogue. It is the converters which are 16, 24, or 32-bit float. So things like the M1 put a mic, preamp, converters, and recorder in one. With separate XLR mics, the capsule is in the mic and all the other things are in the separate recorder.
Love the way you explain and simplify this complicated topic to noobs like me - much appreciated! Is there any software that can take the 32-bit file and do a one-click volume (or db level) normalization? Currently, I am breaking the audio file into chunks and manually adjusting the levels. It's cumbersome and time consuming. Hoping there is a better way!
One problem arising from normalising an entire file is if there is some unnecessary transient which is louder than the rest of the audio - like a mic handling noise for instance - then the wanted part of the file will still be at a low level. Sometimes you have to edit first and normalise quite late in the overall process.
Great video, Curtis! Apparently my Rodecaster Pro 2 is now giving me 32 bit float files, and your video popped into my feed LIKE MAGIC! Any who, I was wondering if this would help me with VO performances where I'm screaming into the mic. (Mostly for video game and other drama based work) Based on what I'm learning here, any help would be minimal, and wouldn't be worth the effort.
Thanks Rob, I’m betting those are just 32-bit float file containers without the fixed gain/multiply converters setup of most 32-bit float recorders. But no, I’d say your quality mics are the most important factor (with their 130dB+ max SPL specs)
Had a buddy record speeches at a wedding with a Zoom F3, but he said even with the DJs gain at the lowest setting, it still distorted heavily. My guess is that he plugged into the back of the speaker which was receiving a line level way too hot for any recorder. I told him that's why I always opt to plug into the microphone receiver if I can, since that's the first gain stage, and is usually too low if anything.
I can summarize the H1 XLR for you: Zoom noisy as expected, small, convenient, slightly larger than the F3, love the combo XLR and the PIP port, screen is super tiny and difficult to read, and the processor is so weak, it takes three times as long to normalize or export as the length of the actual recording. Great for tiny mobile audio interface for live streaming, has loopback.
Very interesting stuff. This reinforces my impression that a Zoom H6 will be more than good enough for me if I want to upgrade further from my H4n Pro. Question not related to 32-bit, but is the Zoom H5 a significant upgrade from the H4n Pro?
@@curtisjudd i know for a fact that it doesn’t :( and that goes for the transmitter (the boom mic/s) and the receiver It’s honestly unfortunate to know that we won’t get your coverage on that product given how interesting it is. They are the only ones making these wireless boom mics targeted at the average everyday creator. I respect your standards, though! Thanks for the reply. I went ahead and ordered a set a while ago for a little below 290 dollars for the dual tx version (retails for 400 in the US but i got it for a bit cheaper during a sale thru a website in my country). Wireless, 32 bit, and internal recording just seems too good to pass up at that price point!
@@skmkenny I understand and I'm sorry about that. I'm just trying to figure out how else to convince product companies to make their products so that we can replace these batteries once they're exhausted. I'm fine with built-in Li-Ion batteries, just asking for a paid kit we can buy to replace it.
@ i’m with you 100%! Hopefully they get on it by the time the vm50 comes out, or better yet some other company can pop up to offer a comparable feature set with user replaceable batteries. Fingers crossed! Thanks for the reply bossman 😎
Indeed, a 32 bit float A/D converter can make a superb recording of the crackle from overloaded mics or preamps in front of it. But I tend to think of it this way - if in the early days of digital recording 32 bit float had been there at the start, would anyone now be extolling the virtues of 24 bit or less and suggesting that we should switch to that? Now we have reached the point where 32 bit float implies no price premium and in effect it is the de facto standard for most gear, even cheap stuff. And maybe some of that cheap stuff is showing the way forward. I have a Zoom M2 mic/recorder in my arsenal and it can be set to normalise on the fly on playback, and produce copies of the recorded files onboard (in an export folder), normalised and output as 24 or 16 bit, ready to use in situations where that can save time later. That device also illustrates how 32 bit float one piece systems have the advantage that the designers can create a level-balanced system, where the mic can handle high levels without clipping, and the preamp can handle the known max level from the mic, and the A/D converters can digitize that well tuned audio with no possible clipping. Maybe that is where 32 bit float really comes into its own. There is also consequent focus in the design on low noise. The other night I did the classic under-the-duvet test of that cheap device and subsequent analysis of the noise indicated that it was just fine for any normal purpose, eg ambient nature recordings or classical music etc (in regard to the self noise). Wow, we have come a long way. The Sony F1/SLF1 system I used for my first digital recording for the Philips label back in 1983 had a 14 bit / 16 bit selector switch. I think I did choose 16 bits...
"Indeed, a 32 bit float A/D converter can make a superb recording of the crackle from overloaded mics or preamps in front of it." Maybe I am misunderstanding the wording, but no bitrate will make a crackling, overloaded mic into a superb recording. The damage is done before it even got the AD converter.
@AdamElteto that's what I'm saying. The converter will record the distorted signal from the preamp. It can't magically fix it. However in a well designed device you would have to work quite hard to cause that overload. It would probably most likely occur when using high output mics close to something remarkably loud.
Great video! However, I think you missed one major benefit of 32 bit float, which is the ability to raise levels in post without raising the device's noise floor if (part of) the audio is too quiet. Of course given that the recording was done with a quality mic and recorder
Thanks 👍 I do find that the noise floor does increase when a clip is digitally boosted in post. But for recordings made on the higher quality recorders, the noise level is still acceptable.
@@OzpeterTrue. But not like a 24bit recording, because a 32bit float recorder uses multiple ADCs. Which, depending on the quality of the recording, makes it possible to raise levels to a certain extent with no noticeable extra noise
@@RedFrameTech I wouldn’t blame you for that. The manufacturers and others have suggested that exact thing. In my practical experience, the noise floor is still there and boosting in post raises it. But what I’ve found is that the noise is generally very much in the acceptable range even after boosting.
I record mostly nature sounds and do a lot of 12-15 hour recordings and one big drawback to 32-bit float only recorders is the files are significantly larger for dynamic range recordings I don't need.I record with my Zoom F3 only when I have to simply because I have to use 32-bit workflow and have much larger file sizes than I really need. Trust me hard drive space fills up quickly on long recordings. Also to everyone who thinks it's a hassle to set gain on a 24-bit recorder then you probably will not want to hear how important microphone placement is and getting your talent or yourself to speak in clear level tones while not turning your head too much etc. The point is good quality audio has so many variables and if you can't handle learning to set the gain then the rest of it is probably going to be too much for you also. So many variables are involved and a 32-bit float recorder is not going to solve all them for you. But I'm sure Zoom,Tascam and the others will want you to think it will be the cure for everything.
Ever record a live rock band? Placement matters....nil. You will have transients pushing 120 db and lows and whispered vocals just barely reaching 50 db in the same performance, no matter where you position your microphones. All you can to is position your mikes the best way you know how to capture the soundstage (or at least make it reasonably correctable in post).
@@chuckschillingvideos Well, you do need to think of the effect of the room in messing with the frequencies of the band, and the balance between audience noise and the band itself.
@@Ozpeter That's accomplished not by mike placement but by orientation and control of the capsule pattern. I have five mikes available to me so it makes it easier to handle the size of the room, how off-axis relative to the PA placement I am, audience size and "enthusiasm" (it matters), and so on. It's not exact, but unless you're working for the venue (not once have I done this), you're usually quite limited in your ability to more your gear around, not to mention you need to be able to keep your eyes on it at all times to protect it from sticky fingers.
I would argue that 32bit float is great for live streaming. I often do live streams of panel discussions. Getting 4-5 Shure sm7b’s into my zoom f6 and pasting it onto reaper via the USB interface mode. Where I do all my dynamics, effects, leveling etc. then passing the “mastered” version onto OBS / Vmix. This allows me to add or remove gain to get to the desired loudness of the platform easily in my daw. If I’m missing something I would love to hear it.
If it works, good 👍 but seems like an extra step to go through reaper to compensate for the fact that audio may exceed 0dB. Why not 24 bit mode with output limiter and processing plugins in OBS?
@@curtisjudd That is very true, depending on the live stream material. I know, that sounds sacrilegious for some audio people, as you should always strive for the best possible quality, but if you have loud sports or spoken words, compressing, limiting and other processing will not significantly detract from the material.
Curtis, interesting video thank you. I have a number of Gongs which I'm interested in recording. I'm a complete novice with respect to sound recording but would suggest that the SPL and dynamic range that a large gong can produce from a main strike (very loud) to the point where the harmonic reverberations drop below audible frequencies is quite wide. I'd be interested in whether you think this would be a valid use case for a 32bit Float capable recorder, not withstanding I'll have to get hold of some equally High Dynamic range microphones to capture the detail at both extremes. Potentially using multiple microphones to capture each end of the frequency range. I'd welcome your thoughts.
Honestly, I think I'd focus mainly on which microphone(s) to use and where to place them. The initial strike may be super loud if miked within a few cm of the gong, but with a gong, I'd think you want to back off a bit because the beauty of its sound is partly the gong and partly how the sound fills the space.
@@curtisjudd Thanks Curtis, I'm going to try microphones located in the same location as the listener about 3-4 feet away. As you say part of the nature of the gong's sound is how it moves the air around the entire room. As a sound therapist I have multiple instruments (gongs, singing bowls, wind chimes, rainsticks & tubular bells) located around the client and I'd like to arrange multiple microphones (4 minimum) possibly two matched pairs, around the client, in order to try & capture the entire session (1hr) as it happens.
very helpful, as always, thank you for this. Can you recommend a good DAW? I've had it with adobe, but not really sure where what other options are generally recommended.
Depends on what type of work you intend to do in the future. If you want to go all-in on post and make that your career? Pro Tools. For most other paths, I’d look at DaVinci Resolve’s Fairlight. You can start with the free version.
@@oriffel Well, if you're using Final Cut Pro, Logic may make sense. Some like Reaper and others Studio One. But in post production for video at the high end, you'll almost always see Pro Tools. Then from there it all depends on workflow and preference. I've transitioned primarily to Fairlight in Resolve.
Finally I'm hearing this. Excellent. This needed to be said. No more 32-bit hype, real deals, gamechangers and other marketing buzzwords. People need to know the actual reality, the advantages as well as shortcomings and caveats. And this is a pretty comprehensive list. Thank you!
It is a game changer if you set gain 20dB higher than optimal like 95% of users do. Even if you have a very good 24bit recorder, the 32bit float recorder will impress the vast majority of users and help them get better results in difficult situations.
@@aristotle_4532I agree it can be a safety net for those doing lots of things on a production, but setting gain is like setting exposure on a camera. Pretty much a fundamental skill of any video production pro or enthusiast.
@@curtisjudd If they can't set gain, can they place a mic? Or in terms of photography, if you can't expose, can you compose? In Greek it makes a better expression because the word for expose also means embarrass in public.
I think the situation where you have to fix the clipping in post is still a big advantage over setting it on the set because you can always make another version of the edit, but you cannot recover wrong gain setting in 24bit file. It's done. However, I see another problem there - as I record media for my small YT channel, I stopped monitoring the sound in any way during the recording process, even with getting a short test record, because I assume the sound can't be "broken". But it can be in many different ways - the Rode Wireless Pro transmitter/recorder can discharge or just go out of order and turn off, the cable from the lavmic or shotgun mic can disconnect or get damaged...
Honestly? I trust NOTHING wireless in recording. It will fail on you at the worst possible time. Be it bluetooth, WIFI, whatever. It will some day set you adrift.
@@chuckschillingvideos I would only do it with a redundant BT/UHF double rig, AND onboard backup recording if it is for recording and not just a live event.
I'd like an updated Sony PCM-A10 that I can easy mount to shoe of mirrorless camera for live music. Are the onboard mics on the Zoom H1essential comparable?
You can probably make that work, but you’re essentially back to working in the space where 24-bit works fine, and encoders will convert to 16 or 24 bit anyway.
what annoys me the most is when reviewers actually tell the audience that 32bit provides higher audio quality, which just proves they have no clue what they are talking about
On some engineering forums where they post measurements, it appears that the distortion performance is worse on some of the popular 32-bit float recorders.
@@curtisjudd I get your point but it's only mainstream "reviewers" who easily just say "better quality" to grab the people's attention. The credible reviewers don't say this as their main point. Also none of them would have mention the mic or audio interface to be the limiting factor like you did, most likely because they aren't even aware of the fact
@@TheNitpicker True, many of us have never had the illusion that 32 automatically means "better". At most, 32 means "flexibility" or "options". But you have to work with those to make things "better quality".
I think where 32-bit really make sense is probably on the processing side. From the best of my super limitted knowledge, some of the DSPs and DAWs are working in 32-bit float internally. They would convert 24-bit integer samples into 32-bit fp first before processing anything, and then convert the output back to 24-bit integer format.
Yes, audio electronics do NOT record audio data in float, which is actually a formula, mantissa ^ exponent. All 32-bit float data is created from a fixed data source, say 24-bit. All 32-bit float eventually ends as fixed data, as Curtis understands about "streaming", because all amplifiers MUST work with real voltages, NOT formulas of what the voltage is (again, m ^ e).
DSP processing, whether in hardware, like a digital mixing console or DSP or software like your DAW, is a very different animal than what is going on with 32 bit float recording. The best and easiest to understand examples I know of are in some of the Yamaha Mixing Consoles manuals. They used to have a really nice graphic that showed how the audio was raised and lowed in the digital space and signal chain to allow for large dynamic ranges that occurred while processing audio and summing together numerous digital audio signals that each could have as much dynamic range as the output of the console is capable of, and what the console is doing internally to handle all that potential.
@@rbingraham If I may quibble ;) Understand you may understand this, but for others. There is NO 32-bit float recording. All original source audio, like microphones, is recorded into a fixed-memory space, either 16 or 24-bit in most electronics. Transposing a recording into 32-bit float requires breaking the value into a mantissa and exponent. 32-bit float is NOT a value, it is a calculation of a value from TWO numbers. The extra 8-bit (from 24-bit) are used to re-scale the data. They do not change the precision of the 24-bit bit precision mantissa. End quibble :)
@@MaxoticsTV That's probably a fair quibble, I don't dive that deeply into what is going on under the hood so to speak. I mostly concern myself with the results that the technology brings and understanding it well enough to use it effectively as possible in my work. My point was mostly just saying that how DAWs handle the math internally (the time from when it leaves the audio driver and is now in the hands of software alone, to when it's summed and sent back to your audio driver) or within a digital console or DSP hardware (which I think mostly use 64 bit floating point? if memory serves) is not really the same as the 32 bit float in your portable audio recorder. Maybe the underlying mathematic fundamentals are more similar than I understand, and that's certainly fully possible. But from the end user perspective and how you use those tools effectively and to their upmost is a fairly different experience. At least in my mind and experience.
I made a video using gh7 w xlr2 w rode ntg 3 in input 2 and rode wireless pro to xlrpro adapter to input 1 i had gain up high on both and set to 20db and imported to fcpx w 48kh quality. Nowhere can i find 32 bit on any file. Would be cool to get some feedback on what im hearing. I knew many questions around i have asked creators and answers are kinda alienating and confusing. If its my mic are those 3 high functioning mics shown in this video a better option? Stereo mic ?idk. The situation was being at a carnival. Tons of noise or sounds. Great breakdown. Thank You
All of those recordings to the GH7 are 24 or 16-bit depending on which recording format you used. The Wireless Pro does not transmit 32-bit float wirelessly. It only records 32-bit float to the transmitters.
@@curtisjudd the gh7 is set to use the xlr2 in camera set to 32 bit 48 in video format mov and prores.i have the ntg3 in the xlr2 port and rode wireless adapted rode vxlr pro to xlr2 port 1 set 48 phantom. I think ive figured it out🤷its the ntg3 . Sound distorts or clippy only through sounds at the side and back of the mic. I connected another vxlr pro to the port the ntg3 was connected and then connected a 3.5 to the vxlr and now both wireles pro and deity mics sound good. As far as the file on the camera card yet to figure out. I really appriciate the feedback! Thank You
I'm even more confused now than before. I have a DJI Mic2 setup and set it to record 32bit float and it records a backup to the TX as well. It actually saved me this past Friday when I found that with 2 mics set up it recorded them as a right and left track, so importing that audio and synching saved me. (Plud I had the bump the audio +20db to match what I usually get) is all this usually possible without the 32bit float? My computer seems to have zero issues.
If it's working for you and you're happy with it, great. I'm just explaining that it isn't the end-all be-all solution for every audio problem and it isn't better audio quality or fidelity, just that you have a fighting chance of recovering the audio if it DOES go over 0dB. That's all.
@curtisjudd Does Auphonics handle 32 bit float files well, or is another step needed before that? I'm a beginner that filmed my first interview yesterday and used Zoom f3. Audio seems fine after listening but a bit low, and nothing clipped. Do I need to do something else with my 32 bit file, or would you think sending it straight to Auphonics would be enough for ok result?
I haven't specifically checked, but I'd be very surprised if Auphonic cannot handle 32-bit float files. And if so, you shouldn't need to do anything beforehand.
please comment about transient issue of 32float, that short tails of noise when a short loud sound ends. looks like it happens because of switching between different stages ADC
@@curtisjudd i saw this on gearspace forum, topic "Problems with 32 bit float for sound design…pitching down and white noise" TH-cam not allow to post link here
Will my dad ever remember how to set gain, or even remember what gain is? No. And that's why he got a Zoom F2 for Christmas. To make it as easy as possible for the guy who still routinely doesn't bother in favor of using his gopro's internal mic, set to point in the wrong direction.
For my use case, i always try my best to set gain, however the performers, sometimes audience, still finds a way to blow up my audio input. So IMHO for me 32 bit all the way. I record live performances I have always set gain "properly" but that 1 section where the audio clips is just not recoverable. 😢 So 32 bit has worked well for my odd use case.
The AT-4021 and 22 it's omni brother supposedly do 146 DB pads off. Pads on I think it's 156 DB SPL. 146 DB SPL is the victory bell siren 100 feet away from it. I think it's a bit more right buy it. Point is I think the 4020 mics can handle that thing. Ah the SE 8 mics again pad on supposedly will do 159 DB. That's mad loud as my favorite favorite little 25 year old next door neighbor Ciara says.
Most problems with distortion is not the 24 bit limits but the dynamic range of the mic pres that is rather bad in all of those recorders under the sound devices level, like zoom and tascam. Also wrong gain staging. Still with the dynamic range of the sound devices and 32bit it is great for FX when using piezos that can be very wild. If not distorted 24bit still sounds better than 32bit
And some of them have poor quality converters amongst some of the budget gear (though less an issue today as it was even 5 years ago). I don’t follow how 24 bit sounds better.
@@curtisjudd It is because 32bit converters are actually two or three 24bit in parallel that split the signal to different level ranges and then they recombine the signals. A kind of a crossover that works in amplitude not frequency domain. In contrast 24bit is a single chip that handles all the conversion without all the above problems. It is mostly theoretical but some people claim they can hear the difference. What is strange is how come and we do not have 32bit single chip converters in widespread use.I think the reason is cause as you said 24bit can theoretically handle 144db already and that is far more than what we need. There are AD converters that offer something close to that but they are hugely expensive cause to achieve that range all the stages including the power supply have to be extremely carefully optimized.
@@peterpeper4837 This may be a cost of 32-bit float. But for spoken word recordings, I haven't been able to point my finger at a specific instance yet where I could confidently say, "That was a 32-bit float artifact." They may be there, I admit that, but I haven't been able to clearly identify such a case for spoken word audio recordings.
So, apologies, but I just want to clarify to make sure I understand one aspect of this. For example, I have the Tascam Portacapture X8. Now, you're saying that if I'm recording in 32bit float, when I set the gain, I'm not altering the preset gain level of the preamp, but I'm adding additional gain to the digital signal further down the chain. So, just so I'm sure I'm crasping this correctly, does that mean that adding gain to the digital signal is further fine-tuning the strength of the signal that's going to be baked in to the recording, is that a good or bad thing if it is, or is it ultimately insignificant since plenty of dynamic range already exists in the recording? Does what I asked make sense?
Maybe it doesn't matter because you would be able to do exactly the same thing in post; doing it in the X8 may just make that step faster or unnecessary.
Yes, but that’s not necessarily a bad thing. It is just math to push the levels back down in post. The potential cost is in less-than-optimally designed preamps with substantial amounts of self noise.
@@curtisjudd I'm not sure if it was my fault in how I structured the question, but I'm not sure if I understood the answer. Easy version, do I bother setting a gain level, or no?
When you set the "pre-amp" gain on a mic you are adjusting the resistance to the voltages coming out of the mic into the pre-amp. You're essentially trying to balance what voltages are more important, the low ones or high ones. Theoretically, we'd want to capture every voltage fluctuation from a mic but in real life we must favor a range of voltages during (pre) amplification. So let's say you gain to get voltages from 1 to 100. But your data space is 1 to 1,000. When you sent those numbers to your headphones it will sound low because it's 1 to 100 at the bottom end of 1 to 1,000. When you increase the volume of the headphones you are simply adding numbers to your recorded numbers. Say 500, so now you have 500 to 600. Great! It sounds at the right volume in your headphones. But maybe you want it louder.... The problem if if you added say 950, then you'd get 950 to 1,050 and 50 of those values would be cut off and you're head distortion--but it's data created distortion. The microphone pre-amp deals with voltages. When you work on the signal you're working on numbers.
Yeah, 32 bit float has a lot of buzz. But all it does really is let us record without having to worry about gain. The "gain" in a 32 bit float recorder is basically just a bit of metadata with a desired playback level. And that was enough for me to move on over to 32bit float for what I do. One thing less to worry about. I then can think more about where the microphone goes and do less fiddling because a meter goes too far sometimes. So, yes. It's not magic. But I am a convert for what it does do for me :)
What drives me crazy is recording someone (speech) and at first they talk really loud or semi loud and then they get really low. I use the old record Tascam dr-40. It does really well on loud sounds if you can set the level right. But when it comes to low sounds. The sound is so low that it is very hard to make it loud enough so it sounds good and you can understand what the person you're recording is saying.
@@Warren1814 Here's one example (though keep in mind that there are more affordable options like Waves Vocal Rider): th-cam.com/video/2svTCBlMIP0/w-d-xo.html
I am a bit confused with how the "dB" is represented in that post-production program (say, Audacity or other) versus the "actual dB". So, human conversation could be 60-70dB (loud,screaming). What do mean by "0 dB" and less, when doing post-production processing, and explaining it? "Recover the sound that goes above 0dB in post-production...", hmm At 10:02 | "Because if it is 32-bit float and it goes above 0, any speaker doesn't matter what kind of speaker it is - is going to distort when you're talking about greater than 0 dB full scale" Those "greater than 0 dB full scale" is really confusing to me to know about :) I mean I imagine the sound volume is going from 0 and above...? NOTE: (I am completely not expert...)
There are different decibel scales. Human conversation occurs in meatspace, where loudness is determined by the air pressure at a given point in space. This is measured in dB SPL (Decibels Sound Pressure Level). Digital audio uses dBFS (Decibels Full-Scale), defined by the number of bits available in a system to represent signal amplitude/headroom (e.g. 16-bit, 24-bit). dBFS has an upper limit called 0 dBFS. If a signal's amplitude exceeds 0 dBFS (i.e. if it contains more bits than are available to the system) then the excessive bits/signal are simply discarded ("clipped") and a characteristic distortion is produced. You can think of it almost like the clearance height of a tunnel. If a rectangular truck with a height of 3.1 meters tries to enter a rectangular tunnel with a clearance of 3.0 meters then 0.1 meters of truck is coming off. Analog signal amplitude is represented by dBv and dBU, which are scales referenced to certain voltages through certain resistances. In analog audio there is no hard clipping per se because the components within the circuit still function when pushed past their nominal operating ranges. Distortion is introduced gradually in response to the amount of overmodulation. Push a little too hard and you get a little distortion. Push very hard and get a lot of distortion.
When I talk about conversation and cite 60-70dB, I’m talking about a range, not absolute dB values. When I talk about going above 0dB I’m referring to dB Full Scale where with 24bit audio, 0dB is the highest amplitude audio possible. Sorry for the confusion and hopefully that helps.
dB is a relative measurement and without a common reference you can’t compare two loudnesses. In a DAW 0 db is typically the maximum loudness that can be exported (over will clip typically) with the negative numbers basically acting like a divider of that maximum loudness.
@@curtisjudd I actually started responding to @jacksonier before I watched the video 😂. I guess I'll watch now and see how badly I misunderstood the question.
Instead of turning down your "clipped" 32-bit signal, why don't you try turning up the 24-bit signal instead? It literally does the exact same thing. It's about 6db dynamic range per bit, so I just boost the signal 18db for 126db DR 21-bit audio. Give me a 64-bit DSP mixer with 24-bit converters please.
@@rbingraham No dude, if you cut a 24-bit signal down to 1/8 the original size it' still 21-bit; it's not possible to have the difference; 21-bit audio is very high quality and digitally boosts to line level perfectly. That is in your head. You're just hearing a higher quality converter. The reality is that most preamps sound really bad a high gain levels, so even if you did turn up the gain you're just capturing noise. There are a lot of other parts in the circuit path, and also there are a lot of 32-bit converters that are not that good. Only some of the ESS 32-bit converters are really good.
That’s great. Now in 2024 can companies start making a file type that have this tech WITH video! I mean a MOV file or MP4 file is essentially a video AND audio file together.. so why not make a new type of file?! I know the size will be bigger but that’s ok with me
I would think yes, as Curtis said, many file formats are just containers of a video and audio track (or multiple tracks). You can try this by combining a video file and a 32-bit float WAV/FLAC track in Handbrake. I would think this file may have playback problems on many current consumer devices.
Well, you might be a sound recording engineer, but it goes really wrong when you try to explain ADC, especially for float. Really ADC doesn't have "errors" anymore as the A/D conversion is linearized with dither noise, and the only factor is the dynamic range set by the number of bits (16-20-24) minus dither noise (+/- 0.5 LSB).
So call dual ADC F32-bit makes things worse. According to ASR's test, performance of Zoom F6 is terrible, and mixpre 2gen also has lower performance than 1gen. Later SD8xx series returned to/insists on traditional 24-bit design.
Some ASR test results: recorders: Zoom F6 SINAD=61db, that means it have 15 or less bit accuracy SD Mixpre 3II SINAD=84db, ~16 bit actual adc accuracy Tascam DR100MK3 SINAD=91db sound interfaces: Motu M4 SINAD=104db, latest Gen 111db Topping 2E2(150$) SINAD 110db RME ADI-2 Pro(1500$) SINAD 117db E1DA ADC(200$ ADC only ) SINAD 119db more ASR test results: microphone preamp noise EIN Topping 2E2 130db Focusrite 2i2 128db RME Babyface Pro 128db SD MixPre3II 110.5db!!!
Reading the comments, it looks like the limiters might have been engaged when reviewing mixpre 3 ii... which is not a great way to measure an ADC performance...
Personally, I have a UAD Apollo X4 and MixPre 6 II. For recording podcast, I am switching to my mixpre 6ii thanks to 32 bits sd card recording while streaming audio to USB. That's the safest you can get when recording long form, and keeping high quality.
32-bit recording is not magic protection from analog overdrive of a/d converter. If in doubt, set it 2dB lower. If really in doubt make it -6. If really really in doubt use the other channel for a safety with a -12dB setting.
I don't know why you didn't contact me. I put a lot of effort in trying to help you. You're very close. I'll have to watch again. Perhaps I'll do a video on it. Anyway, the aspect that blocks you from 100% understanding might be your confusing "scale" for "data". Decibels is a scale. When you talk about audio above 0db, let's say, you're confusing the two. 0db can be scaled anywhere onto the data space. Why would you ever scale your data so something would be above 0 that you don't want above zero? The data (numbers) always come first. Scale comes second. It is a trivial matter in charting/graphs to change scales to anything you want. DBs, logarithms, whatever. The underlying data remains the same. If you can get clear on the difference between the numbers and scale you can move to what the numbers represent. In the case of microphones they represent voltages. All voltages, in all electronics, are limited to a range. Speculating, as you do, about numbers "above 0db" or something like that simply would NOT make sense to an engineer. It's like Spinal Tap. You can write "11" on your guitar amp but that doesn't make it a 11. Anyway, I'm glad you're keeping at it. You're almost there! But you still claim that 32-bit float can help with high dynamic range audio from a microphone. Sorry, that is false. Why do you make claims you cannot prove? You need around 1 bit per 6 dbs. You've looked at a gazillion spec sheets for mics, you do the math ;) 32-bit-float is useful in a studio, sure. Just as it is useful to me in my work with financial numbers. But it is NOT useful in recording microphone data. Indeed, it is an extra layer that adds nothing but more computations for the computer. Indeed, the difference between 24-bit and 32-bit float is SCALE. 32-bit float is 24-bit with an added scale (of exponents). 32-bit float is 24-bit scaled to ^x I'm sure I'm losing you again. Anyone who wants to learn more can google "Microphone Clipping, The Three Types Explained", an essay I wrote on Medium. I explain the 3 types of errors being bandied about by unscrupulous audio equipment manufacturers.
The tone is tough to work with Max, especially when you go make videos suggesting that I’m colluding with manufacturers. Frankly, if your tone was closer to a mentor’s rather than a condescending, “you don’t get it,”, maybe there could be a breakthrough here, if you are correct and there is a breakthrough to be made.
@@curtisjudd I tried to be as polite as possible with you (and everyone). I understand I have bad parts of my personality/communication. I have faults. Please read me back part of any email where I wasn't polite. It's not easy being criticized. I get it. I don't like it myself. In my video I did NOT say you were colluding. I was very careful to say I didn't think you were/are. I said you might be "complicit". It's a completely different thing. You'll never forgive I guess. I've said I'm sorry many times but it takes two to make a breakthrough.
@@MaxoticsTV Let's also be clear: You specifically said that I was lying. That's quite a claim to make. You'd have to know my intent, which you do not. So while you were trying to be polite, it certainly didn't feel that way on my end.
@@curtisjudd I thought about it some more last night and once again I have handled this badly. You are the first TH-camr with a real audience to question the narrative that 32-bit-float fixes clipping or "removes the need for the gain knob." So on that THANK YOU, THANK YOU, THANK YOU. I challenged anyone (you) to investigate the science of it and you did. And let me say again, I don't say do this because I am right, but because I want others to either see what I see, or see what I am missing and explain it to me. Hopefully we all want the same thing--the truth. That said, It's lonely to see one thing and everyone else say it's another because X says it is. If you remember, I said a long time ago it's how these beliefs work through society that greatly interests me. Anyway, I hope we can start over. And thank you again for never giving up on figuring it out, whatever it leads.
I have a 32 bit recorder and used it for an audio recording where the singer overloaded the mic capsule. There's nothing you can do in that situation. For most recordings 32bit float isn't really needed, just nice to have, but other things in your chain will normally have a much bigger impact on clipping.
I do think 32-bit float is mostly hype, but I won't say no to another tool in my toolkit. My suspicion is that manufacturers will start charging extra for 32-bit even though almost no pros will use it or need it in the first place. It reminds me of mics with a hundred different polar patterns despite how almost everyone will just use cardioid. But, the extra polar patterns will still increase the price. I hesitate to say 32-bit is all marketing, but until it's a standard feature on all mics and recorders, it might as well be a gimmick.
There is no need to charge more and with the slew of very cheap 32 bit float devices coming on the market, I don't think that the marketing departments will find it profitable to jack up prices.
I would think such a technology would have to use at some point some limiting or auto-leveling, which would mean you could just use 24 bit in the first place. The converter will not know in advance what the peak will be on that 32-bit stream, so it would have to "react".
You are a pro, and it is second nature for you to set the gain in a recorder. For me -an amateur- setting the gain was my worst anxiety in the entire filming process. Every single time I had a sound problem, it was always because of wrong gain setting. 32-bit float is not marketed for pros. You guys can record amazing audio with 24-bit. For non-professionals, however, 32-bit float IS a game changer because it offers us peace of mind knowing that we will not clip. Great video. As always. Respect.
Yep, this is exactly what ZOOM and Sound Devices like to hear.
On a separate note, I think of setting gain as very much like setting exposure on a camera. Definitely a skill worth developing (and keeping 32-bit float as a safety net).
Another thing many people do is push their gain way too high given the quality preamplifiers we have to work with today. Peaking around -20 is probably what most of us should be aiming for so we'd have far fewer issues with clipping.
You don't need to be a pro to take a minute to set gain properly and to monitor with headphones.
@@rsmith02agreed!
I don't get this, because it seems to me that if you can't figure out how to set gain then you also don't know where you are within your microphone's dynamic range, and _that_ makes it almost impossible to get mic selection and placement right. It seems to me that understanding where your levels are is a prerequisite to any sort of high-quality audio work, not a nice-to-have that's only relevant to the final recording process. After all, you have to avoid saturating your _mic_.
And that's where the marketing for the likes of Zoom really misses the boat. All of their professional-quality sample recordings were undoubtedly created by people who understood exactly where they were within the mic's range and therefore its output voltage.
Not clipping is only one small part of getting good audio, that's only talking about a rookie mistake. There are lots of big mistakes one can make, and that's just the most obvious one that's usually very easily noticed and consequently solved. Taking this one issue off the table doesn't get rid of a host of other potential problems.
By far, the biggest mistake -amateurs- people make is poor microphone placement, including booms, plants, and lavs.
In my area i often have choirs or small music groups that recorded themself, did not level the audio at all or too quiet or too loud, and then i had to mix and repair clipd and super noisy signals ... or i recorded concerts where i gave recorders to the sound guys, just to have a stereo direct feed of the bands on stage, we leveled everything together but later in the evening everything got louder and clipped everything, or at weddings sometimes i have to be super fast and have no time to level again, including going from people talking without a microphone the the audience cheering ... back in the day this ruined so much, but since ichanged to 32bit on everything, this is gone forever. That is why i love it.
Bingo! We're not all recording one or two people talking sitting in a nice studio. :-)
And that's the thing it can help with. 👍
Doing weddings all by yourself will a lot of the times not allow you to check if you're ok with audio levels, and i did a couple of times had the audio levels too high because a priest decided to go way up with his voice mid service and i didn;t notice because i was too busy filming with the cameras. For me, to be able to recover spoken word from some instances, be they as rare as they are, is actually very useful.
👍
Yup, using the right tools for the right circumstances. Nothing wrong with that.
Otherwise, people posting videos on TH-cam really should first study proper film and cinematography and not take the lazy shortcut of digital videography.
Celluloid transfer rules any day over fancy digital schmigital sorcery!
I found 32bit super useful for thunderstorm recordings. Especially when I left my droprigs out in the nature for a several days and I can't adjust gain. It's just perfect for this because I can capture a super quiet atmosphere and when the lightning strikes very close - I just pull that part down in postpro to balance everything without any distortion. I've also found it useful when the wind hits the mics a bit harder. In post-production you just high pass that part and rest of the frequency spectrum will survive without any damage. It's life saver sometimes. But otherwise, for normal ambiences where I not expecting sudden loud noises - 24 bit is more than enough. Btw great explanation as always Curtis!
👍 Keep making great nature recordings!
Important video, finally! I've been emphasizing the clipping potential of most lavs for years, despite the simplified talking point that 32-bit float insures against clipping. Also, it's so important to debunk the oversimplification that a mic will clip "only next to a jet engine." Happens all the time. One last thing to suggest is that tools like De-clip in editors like iZotope RX sometimes repair clipping reasonably well. Also: should have recommended the completely separate tier of lav microphones from the likes of Countryman and DPA that are designed for loud volumes (e.g., singing and sports) sacrificing dynamic range on the quiet end.
Thanks for the notes. 🙏
This may not be the strongest argument. A mic will clip when overdriven regardless of bitrate. At best, the argument can be that SOME people MAY incorrectly think that their mic may not clip under 32-bit.
Thank God you weighed in on this. I did this same topic and got so much unbelievable push back. I don't understand why people PREFER to spend money than just learn how to use their equipment properly
Thanks, John. I'm getting some push back as well. I also suspect that there are some nuanced costs to most 32-bit float implementations but need to do more digging.
Keep making great education over there, friend!
I remember that video. I thought It was one of the best practical presentations on how good the 24-bit tech already is. But I will also say that the answer is already in your question: paying takes less effort than learning, so that's why.
@@mack_solo Sadly, that is probably true for many people.
Appreciate the honesty and in-depth description. I was definitely sold that 32-bit was magic and was the way out of not having to worry about setting levels and fiddling. I'm thankful for your reality check - you'be probably saved me hundreds of hours of future headaches by teaching this.
👍
Recording spoken word in fairly reasonable environments with a MixPre3 II - I use 24bit and do gain staging. Yesterday, as an example, recording "spoken word dialog" of a riverboat ferry captain, on a very windy day + a loud noisy boat I used 32bit float. I still did as much gain staging I could in that unpredictable place as I could and used the fader when needed. Listened to the audio last night and the spoken word came through nice and clear. Nothing of significance to address in post. I probably could have used 24bit, but using 32bit in that situation I at least had the feeling that the margin of error was going to be in my favor.
👍
Still, the most significant factor is that you started with a recorder that has terrific components throughout- great preamps, lots of headroom on the busses, negligible noise and distortion, good metering... ..plus, you rode the faders intelligently.
Excellent video- thank you, Curtis. People overlook that low end 32-bit recorders (ie; Zoom H-series) have nasty built-in mics and preamps. Plus, handling noise must be avoided- it's way worse than any potential benefit from 32-bit float. 32-bit files can be useful, but only if the client knows what to do with them, and we nail all the other factors that create a good recording.
Amen.
Great video! Better understanding the topic. Do you have a video of how to setup and what to think about for the 32 bit float workflow?
Thanks. I have a confession to make: I record almost always in 24-bit. I am pretty decent at setting gain so I'd rather not do extra work in post production with a 32-bit float recorded file that exceeds 0dB. But the main thing you'd have to do for 32-bit float recordings is find those cases where the audio goes over 0dB and either compress them, use clip gain to lower them, or drop the levels of the entire recording so that those bits are below 0dB.
This is SO true, if you don't understand proper gain staging, no amount of 32but float will save you.
👍
I have two quick thoughts:
1. No matter how well you set your gain on any bitrate, even 16 or 24, your microphones can still receive too much pressure. So that alone is not a slam-dunk argument against 32-bit float.
2. For field recordists who set up long-term unattended rigs, they very frequently talk about worrying about an unexpected event/noise clipping their recording.
None of these technologies are wonder drugs. You use the best tools you are able to use for the right purpose, and you learn their ins and outs to apply them properly.
Yes, agreed.
When you are working on your own 32bit float is a godsend. Because recording levels don't have to worried about, you can concentrate on all the other aspects of film making including, microphone placement and all the other thing you mention. I look forward to the time when cameras have 32bit float recording built in.
👍
With 32 bit float, your microphone will probably be the limiting factor. Without 32 bit float it was the recorder. A big improvement!
Assuming there aren’t other costs to 32-bit float. And I’m not 100% sure that’s the case yet.
Biggest takeaways:
1. Master the fundamentals of your craft
2. Use quality equipment (which might not be inexpensive)
3. Repeat
Good wrap up. 👍
This was lots and lots of fun. Crystal-clear, wide-spectrum, HDR coverage (32-bit?) of a neat topic. I enjoyed it from the first frame to the last. Thanks! This is VERY good to know!
👍🙏
Bro…. You don’t miss with content!
@@TheGirlSkateBoardCo thanks 👍
Very good overview of 32 bit float. Maybe the best explanation I have seen so far.
🙏
Thanks for making this video as I learned a lot from your videos as a sound recordist and learning lighting to make me a better boom op! Love your work
🙏
I think you're spot on when talking about recording spoken word and dialog, particularly when working indoors. But we don't always work in doors. And as a recordist, you often do not know when someone might start shouting rather than talking in a normal tone of voice. If it's sudden and you don't know it's coming, it is a game changer in that you now still have workable audio. With a typical 24 bit recording, even if you had the gains set perfectly lets say at -18db FS, it probably clipped at least a little bit. I do agree it's not going to solve all the problems. But it solves a big one if you're working fast and as you said, are using good equipment.
Another point I think you kind of missed is what happens when you have recorded your audio too low or there is a really quiet passage. You sort of touched on it, when talking about quantization errors, but probably should have dived a bit deeper. Having 32 bit float not only means you can bring down audio that went over 0db FS, but it also means you have a lot more bits of resolution at low levels, so if you recorded too softly, then you can boost the audio up without boosting the self noise as much.
My point is, we are not all recording someone or even a few people talking. Some of us record music concerts or ambient sound or all sorts of things that have far more dynamic range than what can be captured with a basic 24 bit recording. And yes it is more than what will be reproduced by the end playback device. But at least we have a better chance of fixing it in post than we did before. I respect your opinion quite a bit, but for certain situations with good gear 32 bit float can be a game changer. It has allowed me to save a lot of recordings that I only had one shot at, that otherwise would have been worthless or very problematic to fix. Some simple volume changes is still a lot faster to fix in post than noise reduction passes and anti clipping passes and all the other tricks we used to do.
One note: I’m not sure that 32-bit float results in lower noise recordings. My experience is that there’s still plenty of ambient noise and self noise depending on the mic and recorder.
A quite educational video, on the aspects of recording technique that you still need to get right in order to benefit from a technology like 32-float recording.
👍
Thank you for that great explaination!
As a consumer recording voice only i just use a used Zoom H1 and set gain lower than -18db. So it never peaks and as i don‘t hear the difference i live with some quantization errors. I don‘t record classical music but voice only, and as long as i am in a quiet room and i don‘t get noise from clothes everything is fine. And all much better than the microphone in the camera or fideling with wireless microphones that may work or may not work. Mixing video and sound from external records is so easy this days with davinci resolve….
👍
Thanks, Curtis, for this very helpful video. I really appreciate it.
Thank you Jay 🙏
Thanks. That is very useful. As a solo-creator / beginner often working in uncontrolled environments, I've gone for 32bit to avoid gain issues. I have a Zoom F2, DJI Mic 2 and a Sennheiser MKE 600 (which I plug into the DJI Mic 2). The problem, however, is low volume for spoken word and there is no way of adjusting this on my equipment. Gain can only be raised in post, where it runs into issues. For example, pushing the gain high enough to make the spoken voice clearly audible often leads to the recording peaking, even though it doesn't sound like it's distorting. It would be great to get a post-production workflow for addressing low volume spoken word on 32bit.
We've got you: th-cam.com/video/lHNhxGojFRU/w-d-xo.html
Or if you want a more automated option: th-cam.com/video/YspZ4gxensI/w-d-xo.html
@@curtisjudd Thanks for the response. The links are very helpful. Very likely to sign up to your Davinci Fairlight course.
Curtis, I'm curious what your limiter settings are on your 888 for spoken-word/dialogue. Threshold, knee, attack, release, ratio? Bus vs. channel? If you could explain what settings you use and why you chose them, that would be really helpful. I think I've got a decent handle on it, but would love to expand my understanding if possible. Thanks!
I have them at the default settings. They rarely engage because I usually leave plenty of headroom for dialogue.
I’ve learnt so much from you and your channel and really appreciate your approach to educating us.
👍
Thank God You are still alive🎉❤ May god bless you
Thanks and you too!
Thanks so much for this fantastic explanation. What a shame though that it'll likely struggle to get the reach of some sponsored influencers sharing rhe marketing hype.
Your knowledge is very much appreciated!
👍
Thanks for this very clear explanation! You constantly refer to "Spoken Word Audio" - what is the typical dynamic range for live musical performances? Thanks again!
I think that depends a lot on the instrument and the music being performed.
You are the best. It could be great to see a review on Zoom M3! nothing much about it really. It's the cheapest 32-bit setup I've seen. I really want to see you a review from you on that product. Thanks a lot
🙏
I've done a lot of recording of individuals speaking in front of a live audience. As only one microphone is used, I would only need to record to one channel. But I always use two and set the output level from the mixer for one of the channels a good bit lower. The higher level channel is set for the "optimum" level and in case of overload, I can replace that short segment using the other channel. I've seen some recorders advertise a feature to automatically record a lower level signal for the same purpose, but I think it's mostly low end equipment that offers that, so I've never tried it. Use your ears, not your eyes. If the waveform display in your DAW or editor looks clipped, listen to it to determine if it's actually objectionable.
As Curtis was saying, the recorder records the signal it's fed, so if it's already distorted, well it is what is, "it" being you failed to use a better mic or you failed to train the speaker or you didn't do whatever it was that could have prevented the distortion.
I've seen that safety track feature, it wasn't on what I'd call low end, it was quite a bit more expensive than entry level pro gear. But it was specifically for one-man-band type setups. It's like those 'composer' synths, they cost 5 times or more what a 'normal good' synth does, but they don't do the same thing. A 'composer synth' you can play a big band setup on by yourself, it's not an instrument, it's a band simulator. It's a drum machine, base synth with sequencer, a synth and multiple lead instruments, all sequencable and live playable at the same time. Probably need a LOT of time to learn how the thing works, or take classes or something. Cool things - sorta, but the multi thousand price tag etc plus steep learning curve is something stopping most from getting into them I think. Except obviously the people employed as musicians using them for 'live' music.
Good strategy.
Excellent as always. Thanks Curtis
👍
9:28 I see the difference of having 32bit vs not having it as (all other things being within spec): "you WILL still have to do the work in post," but at least you CAN still do the work in post.
I'm not certain all other things are equal in every situation, but that's a discussion for a future video.
That was very clear and helpful, thank you.
👍
6:54 The reason we use floating point format is because it allows us to sacrifice precision for range and vice versa - the lower the integral part value, the bigger the fractional part's precision is. When using an integer format the difference between values in quieter parts become relatively smaller and the quantification errors are getting more significant, as you mentioned. Float jest let's us utilise those 4 bytes effectively for sound recording application. But as you said - that's only 1 part of the equation, the hardware needs to be adequate.
💯
As a run a gun team who shoots cars for a living 32 bit recording in the DJI MIC2 is a godsend. We don’t have a dedicated sound guy so I have to set the old recorder (zoom h6) to a very low level in anticipation for the loud engine / exhaust.
I’m surprised that the DJI Mic 2 microphones can handle loud engines and exhaust.
@@curtisjudd I pair them with the Rode Micro and its pretty usable. It does lose out a bit of low end, but I haven't found anything as small & easy to set up with a decent SPL.
Side note, I noticed the 2nd generation of the mic has a far lower Max SPL
@@nazhif1 👍 🏎
Great video Curtis once again. I have a question please. Pro photographers with super wide exposure settings still like to carry a light meter. In pro audio, do you ever use a DB meter when recording louder content?
Fortunately, all of the quality audio recorders have good meters built-in and we use them all the time. 👍
Hi as usual amazing video wonderful information.
may I ask you to make a video if you didn't already about how do you convert 32 bit float audio to 24 bit float in the most common software so that everyone can know how to do such a task as you mentioned that not all post production houses want to get 32 bit float sound so in case we need to convert it
what is the best and fastest way to do it. and with what software.
I would import the clips into whatever DAW app you use, then export the individual clips in wav, 24-bit, keeping the sample rate the same.
I wonder when will they make an 32 bit lav mic capsule, the zoom m1 has a 32bit float shotgun mic but I've only seen like 2 videos that reviewed that mic , so it must not be as great
Mic capsules are analogue. It is the converters which are 16, 24, or 32-bit float. So things like the M1 put a mic, preamp, converters, and recorder in one. With separate XLR mics, the capsule is in the mic and all the other things are in the separate recorder.
Love the way you explain and simplify this complicated topic to noobs like me - much appreciated! Is there any software that can take the 32-bit file and do a one-click volume (or db level) normalization? Currently, I am breaking the audio file into chunks and manually adjusting the levels. It's cumbersome and time consuming. Hoping there is a better way!
DaVinci Resolve. But it doesn’t take care of compressing your audio first, that’s still up to you.
One problem arising from normalising an entire file is if there is some unnecessary transient which is louder than the rest of the audio - like a mic handling noise for instance - then the wanted part of the file will still be at a low level. Sometimes you have to edit first and normalise quite late in the overall process.
@@Ozpeterexactly.
@@curtisjudd Thank you both for chiming in with additional info - much appreciated!
Helpful information. Thanks.
👍
Great video, Curtis! Apparently my Rodecaster Pro 2 is now giving me 32 bit float files, and your video popped into my feed LIKE MAGIC! Any who, I was wondering if this would help me with VO performances where I'm screaming into the mic. (Mostly for video game and other drama based work) Based on what I'm learning here, any help would be minimal, and wouldn't be worth the effort.
Thanks Rob, I’m betting those are just 32-bit float file containers without the fixed gain/multiply converters setup of most 32-bit float recorders.
But no, I’d say your quality mics are the most important factor (with their 130dB+ max SPL specs)
Had a buddy record speeches at a wedding with a Zoom F3, but he said even with the DJs gain at the lowest setting, it still distorted heavily. My guess is that he plugged into the back of the speaker which was receiving a line level way too hot for any recorder. I told him that's why I always opt to plug into the microphone receiver if I can, since that's the first gain stage, and is usually too low if anything.
Or a line level feed direct from the mixing board. But yes, they probably forgot to switch the input on the F3 to line level.
Will you be reviewing the Zoom H1XLR?
Yes
I can summarize the H1 XLR for you: Zoom noisy as expected, small, convenient, slightly larger than the F3, love the combo XLR and the PIP port, screen is super tiny and difficult to read, and the processor is so weak, it takes three times as long to normalize or export as the length of the actual recording. Great for tiny mobile audio interface for live streaming, has loopback.
thank you a lot for this explanation !! 🌟
👍
Very interesting stuff. This reinforces my impression that a Zoom H6 will be more than good enough for me if I want to upgrade further from my H4n Pro. Question not related to 32-bit, but is the Zoom H5 a significant upgrade from the H4n Pro?
Depends on what you need upgraded. The preamps are only very marginally better, probably not going to notice.
@@curtisjudd Thanks Curtis - that was the bit I was wondering about. 🙂
I would be careful investing in the H line, they may be phased out in favor of the H Essential line.
I’m really hoping for you to take a look at the comica vm40! Thanks Curtis
If it has non-user-replaceable batteries, I won't review it. Do you know if it does?
@@curtisjudd I like that bar you set! it is more important than people realize.
@@curtisjudd i know for a fact that it doesn’t :( and that goes for the transmitter (the boom mic/s) and the receiver
It’s honestly unfortunate to know that we won’t get your coverage on that product given how interesting it is. They are the only ones making these wireless boom mics targeted at the average everyday creator. I respect your standards, though! Thanks for the reply.
I went ahead and ordered a set a while ago for a little below 290 dollars for the dual tx version (retails for 400 in the US but i got it for a bit cheaper during a sale thru a website in my country). Wireless, 32 bit, and internal recording just seems too good to pass up at that price point!
@@skmkenny I understand and I'm sorry about that. I'm just trying to figure out how else to convince product companies to make their products so that we can replace these batteries once they're exhausted. I'm fine with built-in Li-Ion batteries, just asking for a paid kit we can buy to replace it.
@ i’m with you 100%! Hopefully they get on it by the time the vm50 comes out, or better yet some other company can pop up to offer a comparable feature set with user replaceable batteries. Fingers crossed! Thanks for the reply bossman 😎
Indeed, a 32 bit float A/D converter can make a superb recording of the crackle from overloaded mics or preamps in front of it. But I tend to think of it this way - if in the early days of digital recording 32 bit float had been there at the start, would anyone now be extolling the virtues of 24 bit or less and suggesting that we should switch to that? Now we have reached the point where 32 bit float implies no price premium and in effect it is the de facto standard for most gear, even cheap stuff.
And maybe some of that cheap stuff is showing the way forward. I have a Zoom M2 mic/recorder in my arsenal and it can be set to normalise on the fly on playback, and produce copies of the recorded files onboard (in an export folder), normalised and output as 24 or 16 bit, ready to use in situations where that can save time later.
That device also illustrates how 32 bit float one piece systems have the advantage that the designers can create a level-balanced system, where the mic can handle high levels without clipping, and the preamp can handle the known max level from the mic, and the A/D converters can digitize that well tuned audio with no possible clipping. Maybe that is where 32 bit float really comes into its own.
There is also consequent focus in the design on low noise. The other night I did the classic under-the-duvet test of that cheap device and subsequent analysis of the noise indicated that it was just fine for any normal purpose, eg ambient nature recordings or classical music etc (in regard to the self noise). Wow, we have come a long way. The Sony F1/SLF1 system I used for my first digital recording for the Philips label back in 1983 had a 14 bit / 16 bit selector switch. I think I did choose 16 bits...
Thanks for sharing
"Indeed, a 32 bit float A/D converter can make a superb recording of the crackle from overloaded mics or preamps in front of it."
Maybe I am misunderstanding the wording, but no bitrate will make a crackling, overloaded mic into a superb recording. The damage is done before it even got the AD converter.
@AdamElteto that's what I'm saying. The converter will record the distorted signal from the preamp. It can't magically fix it. However in a well designed device you would have to work quite hard to cause that overload. It would probably most likely occur when using high output mics close to something remarkably loud.
Great video! However, I think you missed one major benefit of 32 bit float, which is the ability to raise levels in post without raising the device's noise floor if (part of) the audio is too quiet. Of course given that the recording was done with a quality mic and recorder
Any analog noise would be raised in level. Even if it is minimal noise from good analog components.
Thanks 👍 I do find that the noise floor does increase when a clip is digitally boosted in post. But for recordings made on the higher quality recorders, the noise level is still acceptable.
@@OzpeterTrue. But not like a 24bit recording, because a 32bit float recorder uses multiple ADCs. Which, depending on the quality of the recording, makes it possible to raise levels to a certain extent with no noticeable extra noise
@@curtisjuddPerhaps my ‘knowledge’ is incorrect then 😅
@@RedFrameTech I wouldn’t blame you for that. The manufacturers and others have suggested that exact thing. In my practical experience, the noise floor is still there and boosting in post raises it. But what I’ve found is that the noise is generally very much in the acceptable range even after boosting.
I record mostly nature sounds and do a lot of 12-15 hour recordings and one big drawback to 32-bit float only recorders is the files are significantly larger for dynamic range recordings I don't need.I record with my Zoom F3 only when I have to simply because I have to use 32-bit workflow and have much larger file sizes than I really need. Trust me hard drive space fills up quickly on long recordings. Also to everyone who thinks it's a hassle to set gain on a 24-bit recorder then you probably will not want to hear how important microphone placement is and getting your talent or yourself to speak in clear level tones while not turning your head too much etc. The point is good quality audio has so many variables and if you can't handle learning to set the gain then the rest of it is probably going to be too much for you also. So many variables are involved and a 32-bit float recorder is not going to solve all them for you. But I'm sure Zoom,Tascam and the others will want you to think it will be the cure for everything.
Thanks for sharing your experience! I agree that setting gain is a skill worth learning.
Ever record a live rock band? Placement matters....nil. You will have transients pushing 120 db and lows and whispered vocals just barely reaching 50 db in the same performance, no matter where you position your microphones. All you can to is position your mikes the best way you know how to capture the soundstage (or at least make it reasonably correctable in post).
@@chuckschillingvideos Well, you do need to think of the effect of the room in messing with the frequencies of the band, and the balance between audience noise and the band itself.
@@Ozpeter That's accomplished not by mike placement but by orientation and control of the capsule pattern. I have five mikes available to me so it makes it easier to handle the size of the room, how off-axis relative to the PA placement I am, audience size and "enthusiasm" (it matters), and so on. It's not exact, but unless you're working for the venue (not once have I done this), you're usually quite limited in your ability to more your gear around, not to mention you need to be able to keep your eyes on it at all times to protect it from sticky fingers.
I would argue that 32bit float is great for live streaming. I often do live streams of panel discussions. Getting 4-5 Shure sm7b’s into my zoom f6 and pasting it onto reaper via the USB interface mode. Where I do all my dynamics, effects, leveling etc. then passing the “mastered” version onto OBS / Vmix. This allows me to add or remove gain to get to the desired loudness of the platform easily in my daw. If I’m missing something I would love to hear it.
If it works, good 👍 but seems like an extra step to go through reaper to compensate for the fact that audio may exceed 0dB. Why not 24 bit mode with output limiter and processing plugins in OBS?
@@curtisjudd That is very true, depending on the live stream material. I know, that sounds sacrilegious for some audio people, as you should always strive for the best possible quality, but if you have loud sports or spoken words, compressing, limiting and other processing will not significantly detract from the material.
Curtis, interesting video thank you. I have a number of Gongs which I'm interested in recording. I'm a complete novice with respect to sound recording but would suggest that the SPL and dynamic range that a large gong can produce from a main strike (very loud) to the point where the harmonic reverberations drop below audible frequencies is quite wide. I'd be interested in whether you think this would be a valid use case for a 32bit Float capable recorder, not withstanding I'll have to get hold of some equally High Dynamic range microphones to capture the detail at both extremes. Potentially using multiple microphones to capture each end of the frequency range. I'd welcome your thoughts.
Honestly, I think I'd focus mainly on which microphone(s) to use and where to place them. The initial strike may be super loud if miked within a few cm of the gong, but with a gong, I'd think you want to back off a bit because the beauty of its sound is partly the gong and partly how the sound fills the space.
@@curtisjudd Thanks Curtis, I'm going to try microphones located in the same location as the listener about 3-4 feet away. As you say part of the nature of the gong's sound is how it moves the air around the entire room. As a sound therapist I have multiple instruments (gongs, singing bowls, wind chimes, rainsticks & tubular bells) located around the client and I'd like to arrange multiple microphones (4 minimum) possibly two matched pairs, around the client, in order to try & capture the entire session (1hr) as it happens.
very helpful, as always, thank you for this.
Can you recommend a good DAW? I've had it with adobe, but not really sure where what other options are generally recommended.
Depends on what type of work you intend to do in the future. If you want to go all-in on post and make that your career? Pro Tools. For most other paths, I’d look at DaVinci Resolve’s Fairlight. You can start with the free version.
@@curtisjudd i do mostly video work, corporate usually internal stuff, been using da vinchi for about a year, jut curious what else is out there.
@@oriffel Well, if you're using Final Cut Pro, Logic may make sense. Some like Reaper and others Studio One. But in post production for video at the high end, you'll almost always see Pro Tools. Then from there it all depends on workflow and preference. I've transitioned primarily to Fairlight in Resolve.
Finally I'm hearing this. Excellent. This needed to be said. No more 32-bit hype, real deals, gamechangers and other marketing buzzwords. People need to know the actual reality, the advantages as well as shortcomings and caveats. And this is a pretty comprehensive list. Thank you!
Exsctly...its an amateur gimmick not a new do it all
👍
It is a game changer if you set gain 20dB higher than optimal like 95% of users do. Even if you have a very good 24bit recorder, the 32bit float recorder will impress the vast majority of users and help them get better results in difficult situations.
@@aristotle_4532I agree it can be a safety net for those doing lots of things on a production, but setting gain is like setting exposure on a camera. Pretty much a fundamental skill of any video production pro or enthusiast.
@@curtisjudd If they can't set gain, can they place a mic? Or in terms of photography, if you can't expose, can you compose? In Greek it makes a better expression because the word for expose also means embarrass in public.
Love your tough love Curtis :)
❤️
Brilliant, thank you so much
👍
I think the situation where you have to fix the clipping in post is still a big advantage over setting it on the set because you can always make another version of the edit, but you cannot recover wrong gain setting in 24bit file. It's done.
However, I see another problem there - as I record media for my small YT channel, I stopped monitoring the sound in any way during the recording process, even with getting a short test record, because I assume the sound can't be "broken". But it can be in many different ways - the Rode Wireless Pro transmitter/recorder can discharge or just go out of order and turn off, the cable from the lavmic or shotgun mic can disconnect or get damaged...
Honestly? I trust NOTHING wireless in recording. It will fail on you at the worst possible time. Be it bluetooth, WIFI, whatever. It will some day set you adrift.
Agreed, always monitor when relying on wireless. There WILL be problems at some point.
@@chuckschillingvideos I would only do it with a redundant BT/UHF double rig, AND onboard backup recording if it is for recording and not just a live event.
I'd like an updated Sony PCM-A10 that I can easy mount to shoe of mirrorless camera for live music. Are the onboard mics on the Zoom H1essential comparable?
I haven't tested the Sony recorder, but I'd doubt the the H1's mics are better.
Super-informative, Curtis. I will be passing this along the to the Ecamm community who are commonly afflicted with G.A.S. 🤣
LOL
About 32 bit float in live streaming, a digital limiter should work quite well as there is no digital clipping in 32 bit float.
You can probably make that work, but you’re essentially back to working in the space where 24-bit works fine, and encoders will convert to 16 or 24 bit anyway.
Excellent ! Thank you
👍
Well explained.
🙏
what annoys me the most is when reviewers actually tell the audience that 32bit provides higher audio quality, which just proves they have no clue what they are talking about
Well, if it does avoid distortion then tne quality will be higher. But I see your point.
On some engineering forums where they post measurements, it appears that the distortion performance is worse on some of the popular 32-bit float recorders.
@@curtisjudd I get your point but it's only mainstream "reviewers" who easily just say "better quality" to grab the people's attention. The credible reviewers don't say this as their main point. Also none of them would have mention the mic or audio interface to be the limiting factor like you did, most likely because they aren't even aware of the fact
Fair point @TheNitpicker
@@TheNitpicker True, many of us have never had the illusion that 32 automatically means "better". At most, 32 means "flexibility" or "options". But you have to work with those to make things "better quality".
Thank you!
👍
I think where 32-bit really make sense is probably on the processing side. From the best of my super limitted knowledge, some of the DSPs and DAWs are working in 32-bit float internally. They would convert 24-bit integer samples into 32-bit fp first before processing anything, and then convert the output back to 24-bit integer format.
64-bit internally, in some cases.
Yes, audio electronics do NOT record audio data in float, which is actually a formula, mantissa ^ exponent. All 32-bit float data is created from a fixed data source, say 24-bit. All 32-bit float eventually ends as fixed data, as Curtis understands about "streaming", because all amplifiers MUST work with real voltages, NOT formulas of what the voltage is (again, m ^ e).
DSP processing, whether in hardware, like a digital mixing console or DSP or software like your DAW, is a very different animal than what is going on with 32 bit float recording. The best and easiest to understand examples I know of are in some of the Yamaha Mixing Consoles manuals. They used to have a really nice graphic that showed how the audio was raised and lowed in the digital space and signal chain to allow for large dynamic ranges that occurred while processing audio and summing together numerous digital audio signals that each could have as much dynamic range as the output of the console is capable of, and what the console is doing internally to handle all that potential.
@@rbingraham If I may quibble ;) Understand you may understand this, but for others. There is NO 32-bit float recording. All original source audio, like microphones, is recorded into a fixed-memory space, either 16 or 24-bit in most electronics. Transposing a recording into 32-bit float requires breaking the value into a mantissa and exponent. 32-bit float is NOT a value, it is a calculation of a value from TWO numbers. The extra 8-bit (from 24-bit) are used to re-scale the data. They do not change the precision of the 24-bit bit precision mantissa. End quibble :)
@@MaxoticsTV That's probably a fair quibble, I don't dive that deeply into what is going on under the hood so to speak. I mostly concern myself with the results that the technology brings and understanding it well enough to use it effectively as possible in my work. My point was mostly just saying that how DAWs handle the math internally (the time from when it leaves the audio driver and is now in the hands of software alone, to when it's summed and sent back to your audio driver) or within a digital console or DSP hardware (which I think mostly use 64 bit floating point? if memory serves) is not really the same as the 32 bit float in your portable audio recorder. Maybe the underlying mathematic fundamentals are more similar than I understand, and that's certainly fully possible. But from the end user perspective and how you use those tools effectively and to their upmost is a fairly different experience. At least in my mind and experience.
I made a video using gh7 w xlr2 w rode ntg 3 in input 2 and rode wireless pro to xlrpro adapter to input 1 i had gain up high on both and set to 20db and imported to fcpx w 48kh quality. Nowhere can i find 32 bit on any file. Would be cool to get some feedback on what im hearing. I knew many questions around i have asked creators and answers are kinda alienating and confusing. If its my mic are those 3 high functioning mics shown in this video a better option? Stereo mic ?idk. The situation was being at a carnival. Tons of noise or sounds. Great breakdown. Thank You
All of those recordings to the GH7 are 24 or 16-bit depending on which recording format you used. The Wireless Pro does not transmit 32-bit float wirelessly. It only records 32-bit float to the transmitters.
@@curtisjudd the gh7 is set to use the xlr2 in camera set to 32 bit 48 in video format mov and prores.i have the ntg3 in the xlr2 port and rode wireless adapted rode vxlr pro to xlr2 port 1 set 48 phantom. I think ive figured it out🤷its the ntg3 . Sound distorts or clippy only through sounds at the side and back of the mic. I connected another vxlr pro to the port the ntg3 was connected and then connected a 3.5 to the vxlr and now both wireles pro and deity mics sound good. As far as the file on the camera card yet to figure out. I really appriciate the feedback! Thank You
@@DroseMr 👍
I'm even more confused now than before. I have a DJI Mic2 setup and set it to record 32bit float and it records a backup to the TX as well. It actually saved me this past Friday when I found that with 2 mics set up it recorded them as a right and left track, so importing that audio and synching saved me. (Plud I had the bump the audio +20db to match what I usually get) is all this usually possible without the 32bit float? My computer seems to have zero issues.
If it's working for you and you're happy with it, great. I'm just explaining that it isn't the end-all be-all solution for every audio problem and it isn't better audio quality or fidelity, just that you have a fighting chance of recovering the audio if it DOES go over 0dB. That's all.
@curtisjudd Does Auphonics handle 32 bit float files well, or is another step needed before that? I'm a beginner that filmed my first interview yesterday and used Zoom f3. Audio seems fine after listening but a bit low, and nothing clipped. Do I need to do something else with my 32 bit file, or would you think sending it straight to Auphonics would be enough for ok result?
I haven't specifically checked, but I'd be very surprised if Auphonic cannot handle 32-bit float files. And if so, you shouldn't need to do anything beforehand.
@@curtisjudd Great. Thanks for taking the time to answer questions here!
please comment about transient issue of 32float, that short tails of noise when a short loud sound ends. looks like it happens because of switching between different stages ADC
I’d love to learn more about this. Any recommendations on where to learn more?
@@curtisjudd i saw this on gearspace forum, topic "Problems with 32 bit float for sound design…pitching down and white noise"
TH-cam not allow to post link here
@@alexandrzatonskyok, thanks.
Sound pressure level of your mic is crucial
Yes. 👍
Will my dad ever remember how to set gain, or even remember what gain is?
No.
And that's why he got a Zoom F2 for Christmas. To make it as easy as possible for the guy who still routinely doesn't bother in favor of using his gopro's internal mic, set to point in the wrong direction.
Hahaha, yes! But then who manages the audio in post when he comes back with audio that nobody can hear? 😉
For my use case, i always try my best to set gain, however the performers, sometimes audience, still finds a way to blow up my audio input. So IMHO for me 32 bit all the way. I record live performances I have always set gain "properly" but that 1 section where the audio clips is just not recoverable. 😢
So 32 bit has worked well for my odd use case.
Ok 👍
What are the specs for range in Deity PR2! Not in their specs.
Not sure on the PR-2, but the included W-Lav Pro is 110dB per the specs.
The summery:
“32 bit float is not magical. It’s just a clever engineering.”
Thanks as always, Curtis 👍🏾
👍
The AT-4021 and 22 it's omni brother supposedly do 146 DB pads off.
Pads on I think it's 156 DB SPL.
146 DB SPL is the victory bell siren 100 feet away from it. I think it's a bit more right buy it. Point is I think the 4020 mics can handle that thing.
Ah the SE 8 mics again pad on supposedly will do 159 DB. That's mad loud as my favorite favorite little 25 year old next door neighbor Ciara says.
Epic microphones.
@@curtisjudd Hell to the yeah.
Now I have an F6 I really should look into those.
Is the zoom h6 a good 32 bit recorder?
It is a fine recorder, but the H6 does not record to 32-bit float. The H6 Essential does.
@ I got the f3 yesterday!
@@directorparisholmes2024 Happy recording!
Most problems with distortion is not the 24 bit limits but the dynamic range of the mic pres that is rather bad in all of those recorders under the sound devices level, like zoom and tascam. Also wrong gain staging. Still with the dynamic range of the sound devices and 32bit it is great for FX when using piezos that can be very wild.
If not distorted 24bit still sounds better than 32bit
And some of them have poor quality converters amongst some of the budget gear (though less an issue today as it was even 5 years ago).
I don’t follow how 24 bit sounds better.
@@curtisjudd It is because 32bit converters are actually two or three 24bit in parallel that split the signal to different level ranges and then they recombine the signals. A kind of a crossover that works in amplitude not frequency domain.
In contrast 24bit is a single chip that handles all the conversion without all the above problems. It is mostly theoretical but some people claim they can hear the difference. What is strange is how come and we do not have 32bit single chip converters in widespread use.I think the reason is cause as you said 24bit can theoretically handle 144db already and that is far more than what we need. There are AD converters that offer something close to that but they are hugely expensive cause to achieve that range all the stages including the power supply have to be extremely carefully optimized.
@@peterpeper4837 This may be a cost of 32-bit float. But for spoken word recordings, I haven't been able to point my finger at a specific instance yet where I could confidently say, "That was a 32-bit float artifact." They may be there, I admit that, but I haven't been able to clearly identify such a case for spoken word audio recordings.
So, apologies, but I just want to clarify to make sure I understand one aspect of this. For example, I have the Tascam Portacapture X8. Now, you're saying that if I'm recording in 32bit float, when I set the gain, I'm not altering the preset gain level of the preamp, but I'm adding additional gain to the digital signal further down the chain. So, just so I'm sure I'm crasping this correctly, does that mean that adding gain to the digital signal is further fine-tuning the strength of the signal that's going to be baked in to the recording, is that a good or bad thing if it is, or is it ultimately insignificant since plenty of dynamic range already exists in the recording? Does what I asked make sense?
Maybe it doesn't matter because you would be able to do exactly the same thing in post; doing it in the X8 may just make that step faster or unnecessary.
Yes, but that’s not necessarily a bad thing. It is just math to push the levels back down in post. The potential cost is in less-than-optimally designed preamps with substantial amounts of self noise.
@@curtisjudd I'm not sure if it was my fault in how I structured the question, but I'm not sure if I understood the answer. Easy version, do I bother setting a gain level, or no?
When you set the "pre-amp" gain on a mic you are adjusting the resistance to the voltages coming out of the mic into the pre-amp. You're essentially trying to balance what voltages are more important, the low ones or high ones. Theoretically, we'd want to capture every voltage fluctuation from a mic but in real life we must favor a range of voltages during (pre) amplification.
So let's say you gain to get voltages from 1 to 100. But your data space is 1 to 1,000. When you sent those numbers to your headphones it will sound low because it's 1 to 100 at the bottom end of 1 to 1,000. When you increase the volume of the headphones you are simply adding numbers to your recorded numbers. Say 500, so now you have 500 to 600. Great! It sounds at the right volume in your headphones. But maybe you want it louder....
The problem if if you added say 950, then you'd get 950 to 1,050 and 50 of those values would be cut off and you're head distortion--but it's data created distortion. The microphone pre-amp deals with voltages. When you work on the signal you're working on numbers.
@@MaxoticsTV ok. I'm not sure how that relates to my question, but thank you.
Yeah, 32 bit float has a lot of buzz. But all it does really is let us record without having to worry about gain. The "gain" in a 32 bit float recorder is basically just a bit of metadata with a desired playback level.
And that was enough for me to move on over to 32bit float for what I do. One thing less to worry about. I then can think more about where the microphone goes and do less fiddling because a meter goes too far sometimes.
So, yes. It's not magic. But I am a convert for what it does do for me :)
With a good microphone, 32-bit float should serve spoken word recordings quite well.
wow 32bit recording demystified.
👍
What drives me crazy is recording someone (speech) and at first they talk really loud or semi loud and then they get really low. I use the old record Tascam dr-40. It does really well on loud sounds if you can set the level right. But when it comes to low sounds. The sound is so low that it is very hard to make it loud enough so it sounds good and you can understand what the person you're recording is saying.
32-bit float doesn't even this out for you, just to be clear.
@@curtisjudd Is there anything that can?
@@Warren1814levelers are made to do that.
@@curtisjudd Oh, oh ok... i am new to all of this.. so i will have to look it up and see what one looks like.
@@Warren1814 Here's one example (though keep in mind that there are more affordable options like Waves Vocal Rider): th-cam.com/video/2svTCBlMIP0/w-d-xo.html
Neumann kms 104p, stage mike, max spl at 0.5 % distortion, 150db. Main mike for my podcast, awesome sound
👍
Now let’s hear Paul Allen’s 24-bit file recorded at -400dB.
And with 32-bit float as well. Worth a comparison.
Rode K2. 152dB of DR.
Wow.
I am a bit confused with how the "dB" is represented in that post-production program (say, Audacity or other) versus the "actual dB". So, human conversation could be 60-70dB (loud,screaming). What do mean by "0 dB" and less, when doing post-production processing, and explaining it?
"Recover the sound that goes above 0dB in post-production...", hmm
At 10:02 | "Because if it is 32-bit float and it goes above 0, any speaker doesn't matter what kind of speaker it is - is going to distort when you're talking about greater than 0 dB full scale"
Those "greater than 0 dB full scale" is really confusing to me to know about :) I mean I imagine the sound volume is going from 0 and above...?
NOTE: (I am completely not expert...)
There are different decibel scales. Human conversation occurs in meatspace, where loudness is determined by the air pressure at a given point in space. This is measured in dB SPL (Decibels Sound Pressure Level).
Digital audio uses dBFS (Decibels Full-Scale), defined by the number of bits available in a system to represent signal amplitude/headroom (e.g. 16-bit, 24-bit). dBFS has an upper limit called 0 dBFS. If a signal's amplitude exceeds 0 dBFS (i.e. if it contains more bits than are available to the system) then the excessive bits/signal are simply discarded ("clipped") and a characteristic distortion is produced.
You can think of it almost like the clearance height of a tunnel. If a rectangular truck with a height of 3.1 meters tries to enter a rectangular tunnel with a clearance of 3.0 meters then 0.1 meters of truck is coming off.
Analog signal amplitude is represented by dBv and dBU, which are scales referenced to certain voltages through certain resistances. In analog audio there is no hard clipping per se because the components within the circuit still function when pushed past their nominal operating ranges. Distortion is introduced gradually in response to the amount of overmodulation. Push a little too hard and you get a little distortion. Push very hard and get a lot of distortion.
When I talk about conversation and cite 60-70dB, I’m talking about a range, not absolute dB values.
When I talk about going above 0dB I’m referring to dB Full Scale where with 24bit audio, 0dB is the highest amplitude audio possible.
Sorry for the confusion and hopefully that helps.
dB is a relative measurement and without a common reference you can’t compare two loudnesses. In a DAW 0 db is typically the maximum loudness that can be exported (over will clip typically) with the negative numbers basically acting like a divider of that maximum loudness.
Thank you very much all for the explanations!) I will review them more carefully.
@@curtisjudd I actually started responding to @jacksonier before I watched the video 😂. I guess I'll watch now and see how badly I misunderstood the question.
Instead of turning down your "clipped" 32-bit signal, why don't you try turning up the 24-bit signal instead? It literally does the exact same thing. It's about 6db dynamic range per bit, so I just boost the signal 18db for 126db DR 21-bit audio. Give me a 64-bit DSP mixer with 24-bit converters please.
I record in 24-bit 99.9% of the time. I'm with you.
In my experience boosting soft levels in 24 bit does not get you the same results as it does in 32 bit. To each their own though.
@@rbingraham No dude, if you cut a 24-bit signal down to 1/8 the original size it' still 21-bit; it's not possible to have the difference; 21-bit audio is very high quality and digitally boosts to line level perfectly. That is in your head. You're just hearing a higher quality converter. The reality is that most preamps sound really bad a high gain levels, so even if you did turn up the gain you're just capturing noise. There are a lot of other parts in the circuit path, and also there are a lot of 32-bit converters that are not that good. Only some of the ESS 32-bit converters are really good.
That’s great. Now in 2024 can companies start making a file type that have this tech WITH video! I mean a MOV file or MP4 file is essentially a video AND audio file together.. so why not make a new type of file?! I know the size will be bigger but that’s ok with me
Ah, I wonder if that's already possible with some file container types...
I would think yes, as Curtis said, many file formats are just containers of a video and audio track (or multiple tracks).
You can try this by combining a video file and a 32-bit float WAV/FLAC track in Handbrake.
I would think this file may have playback problems on many current consumer devices.
Thank you for the much needed reality check!
👍
The Rode video mic X has a SPL of 184db from memory. So....
Nice
I had the zoom f2. It worked for a couple of months then stopped working. Don’t waste your money.
Uh oh.
Float is not magic. But it still has its function.
Yes.
Well, you might be a sound recording engineer, but it goes really wrong when you try to explain ADC, especially for float. Really ADC doesn't have "errors" anymore as the A/D conversion is linearized with dither noise, and the only factor is the dynamic range set by the number of bits (16-20-24) minus dither noise (+/- 0.5 LSB).
Ok, thanks for the correction.
So call dual ADC F32-bit makes things worse. According to ASR's test, performance of Zoom F6 is terrible, and mixpre 2gen also has lower performance than 1gen. Later SD8xx series returned to/insists on traditional 24-bit design.
Some ASR test results:
recorders:
Zoom F6 SINAD=61db, that means it have 15 or less bit accuracy
SD Mixpre 3II SINAD=84db, ~16 bit actual adc accuracy
Tascam DR100MK3 SINAD=91db
sound interfaces:
Motu M4 SINAD=104db, latest Gen 111db
Topping 2E2(150$) SINAD 110db
RME ADI-2 Pro(1500$) SINAD 117db
E1DA ADC(200$ ADC only ) SINAD 119db
more ASR test results: microphone preamp noise EIN
Topping 2E2 130db
Focusrite 2i2 128db
RME Babyface Pro 128db
SD MixPre3II 110.5db!!!
Where are these tests published, please?
@@curtisjudd ASR-audio science review
Reading the comments, it looks like the limiters might have been engaged when reviewing mixpre 3 ii... which is not a great way to measure an ADC performance...
Personally, I have a UAD Apollo X4 and MixPre 6 II. For recording podcast, I am switching to my mixpre 6ii thanks to 32 bits sd card recording while streaming audio to USB. That's the safest you can get when recording long form, and keeping high quality.
32-bit recording is not magic protection from analog overdrive of a/d converter. If in doubt, set it 2dB lower. If really in doubt make it -6. If really really in doubt use the other channel for a safety with a -12dB setting.
Good call.
I don't know why you didn't contact me. I put a lot of effort in trying to help you. You're very close. I'll have to watch again. Perhaps I'll do a video on it. Anyway, the aspect that blocks you from 100% understanding might be your confusing "scale" for "data". Decibels is a scale. When you talk about audio above 0db, let's say, you're confusing the two. 0db can be scaled anywhere onto the data space. Why would you ever scale your data so something would be above 0 that you don't want above zero?
The data (numbers) always come first. Scale comes second. It is a trivial matter in charting/graphs to change scales to anything you want. DBs, logarithms, whatever. The underlying data remains the same.
If you can get clear on the difference between the numbers and scale you can move to what the numbers represent. In the case of microphones they represent voltages. All voltages, in all electronics, are limited to a range. Speculating, as you do, about numbers "above 0db" or something like that simply would NOT make sense to an engineer. It's like Spinal Tap. You can write "11" on your guitar amp but that doesn't make it a 11.
Anyway, I'm glad you're keeping at it. You're almost there! But you still claim that 32-bit float can help with high dynamic range audio from a microphone. Sorry, that is false. Why do you make claims you cannot prove? You need around 1 bit per 6 dbs. You've looked at a gazillion spec sheets for mics, you do the math ;)
32-bit-float is useful in a studio, sure. Just as it is useful to me in my work with financial numbers. But it is NOT useful in recording microphone data. Indeed, it is an extra layer that adds nothing but more computations for the computer.
Indeed, the difference between 24-bit and 32-bit float is SCALE. 32-bit float is 24-bit with an added scale (of exponents). 32-bit float is 24-bit scaled to ^x I'm sure I'm losing you again. Anyone who wants to learn more can google "Microphone Clipping, The Three Types Explained", an essay I wrote on Medium. I explain the 3 types of errors being bandied about by unscrupulous audio equipment manufacturers.
The tone is tough to work with Max, especially when you go make videos suggesting that I’m colluding with manufacturers. Frankly, if your tone was closer to a mentor’s rather than a condescending, “you don’t get it,”, maybe there could be a breakthrough here, if you are correct and there is a breakthrough to be made.
@@curtisjudd I tried to be as polite as possible with you (and everyone). I understand I have bad parts of my personality/communication. I have faults. Please read me back part of any email where I wasn't polite. It's not easy being criticized. I get it. I don't like it myself. In my video I did NOT say you were colluding. I was very careful to say I didn't think you were/are. I said you might be "complicit". It's a completely different thing. You'll never forgive I guess. I've said I'm sorry many times but it takes two to make a breakthrough.
@@MaxoticsTV Then please, try explaining again. Last time it was beyond my ability to understand, at least the way you described it to me.
@@MaxoticsTV Let's also be clear: You specifically said that I was lying. That's quite a claim to make. You'd have to know my intent, which you do not. So while you were trying to be polite, it certainly didn't feel that way on my end.
@@curtisjudd I thought about it some more last night and once again I have handled this badly. You are the first TH-camr with a real audience to question the narrative that 32-bit-float fixes clipping or "removes the need for the gain knob." So on that THANK YOU, THANK YOU, THANK YOU. I challenged anyone (you) to investigate the science of it and you did. And let me say again, I don't say do this because I am right, but because I want others to either see what I see, or see what I am missing and explain it to me. Hopefully we all want the same thing--the truth.
That said, It's lonely to see one thing and everyone else say it's another because X says it is. If you remember, I said a long time ago it's how these beliefs work through society that greatly interests me. Anyway, I hope we can start over. And thank you again for never giving up on figuring it out, whatever it leads.
I have a 32 bit recorder and used it for an audio recording where the singer overloaded the mic capsule. There's nothing you can do in that situation. For most recordings 32bit float isn't really needed, just nice to have, but other things in your chain will normally have a much bigger impact on clipping.
👍
I do think 32-bit float is mostly hype, but I won't say no to another tool in my toolkit. My suspicion is that manufacturers will start charging extra for 32-bit even though almost no pros will use it or need it in the first place. It reminds me of mics with a hundred different polar patterns despite how almost everyone will just use cardioid. But, the extra polar patterns will still increase the price. I hesitate to say 32-bit is all marketing, but until it's a standard feature on all mics and recorders, it might as well be a gimmick.
Yes, for video dialogue we rarely need more than one of the cardioid polar patterns.
There is no need to charge more and with the slew of very cheap 32 bit float devices coming on the market, I don't think that the marketing departments will find it profitable to jack up prices.
Just the opposite really, it keeps getting more and more affordable as the years pass.
I'm sure someone's working on a real-time 32-bit to 24-bit plugin for live audio.
Interesting. I wonder what problem they’re aiming to solve with such a thing. Why not just use 24-bit from the start, in other words.
I would think such a technology would have to use at some point some limiting or auto-leveling, which would mean you could just use 24 bit in the first place. The converter will not know in advance what the peak will be on that 32-bit stream, so it would have to "react".
"Yes" - Cue the Outro. Are you angling for Microphone sponsorships?
Nope, not interested in sponsorships.