Game-changer: How 32-Bit Float Recording Makes Clipping a Thing of the Past 

Missed a great opportunity to state that "32-bit float recording is a complete GAIN changer" 🤣

I think HDR photography would be a much better analogy. In audio, you have this window you have to shift around. In photography that window is the dynamic range of the camera sensor and it comes exactly with the same problem. When you don't set that window to the correct setting (gain or exposure settings in the camera) some samples/pixel will be outside of that range (clipping/hissing/non-audible or over or under exposured) which can't be restored in post because information was lost. Sometimes, what you want to capture has such a big range in volume/brightness that the window you're shifting around just isn't big enough (very loud AND quite impulses or very bright AND dark areas in the picture). To deal with that you just make multiple recordings (at once with audio/in a row with photo) and then combine them by having different pairs of preamps with different gain settings and A/D converters, choosing the converter which is within that window for that specific sample or taking multiple photos and choosing the photo with a good exposure for that specific pixel. RAW or JPEG files are a better analogy to WAV or MP3 files. JPEG and MP3 files are compressed and information was lost. RAW and WAV both are uncompressed and all the information of the recording is still there. But just with a RAW photo without HDR, you can still have the problem of over- AND underexposed pixels when the dynamic range of the sensor is to low for that picture.

This is a great video explaining how 32-bit float recording works! I think the only aspect that's missing is the notion that you still need to practice good mic technique and placement when recording in 32-bit float. An analogy I like to use is to think of 32-bit float recording like shooting a 360 video. With a 360 video, you can simply hit record and reframe your shot in post, much like 32-bit float recording and the ability to "reframe" the gain in post. However, if your initial 360 camera position is, for instance, too low, you can't magically change the perspective. Likewise, if your mic is too far away from your subject, and you're in a noisy environment, raising the gain in post brings up not only the voice being recorded, but also the ambient environmental sounds. Conversely, if you're too close to the mic, plosives will still ruin a 32-bit float recording.

This is so important! I really like how Dan's strategy to teach 32-bit float is to begin by explaining why DAWs use it -- that's the best way to see how it especially makes sense there, and why it's where it's been for the longest. If this educational production portends a whole 32-bit float product line on the way from Rode, this is truly exciting! Zoom has phoned it in recently after their fantastic F6, with poorly engineered MicTrak products, and Rode has a great opening here.

For the math-curious, the "float" concept is basically just scientific notation, such as 1.2345x10^-6, where most of the 32 bits are used to store the "mantissa" (1.2345..) and a few store the exponent (-6). This lets you represent a vast range of values without losing important details. What's novel and exciting (to me) is that Rode integrated an ultra-wide dynamic range 32-bit float A-to-D converter into the already lowest-noise-ever NT1 that should now allow you to capture the artist's eyes blinking from across the room. One might say it's a complete GAIN changer :D

In this video you guys are forgetting that preamplification happens before conversion, in the analog realm. 0VU is still 0VU, no matter how much digital headroom you have. So what you are talking about is still processing and not recording.

32 bits is really more like HDR photography based on muti exposure than Raw, it is about passing the sound through multiple amplifiers at different stage (just like HDR photography does with different levels of image exposure). Raw means no data compression, which is what wavefiles already are. So I dont think the analogy is the best here. Also, most DAW do make pre-render in 32 bits but can calculate at 64 bits precision.

As a camera guy, the raw comparison was a perfect analogy. Thank you.

Great resource. Thank you, RØDE!

Thank you Dan, learned a ton today. Sounds :-) like 32-bit float is a game changer. I’ll try and dig into this.

is the gain knob obsolete or does it define how the preamp colors sounds ?

Thanks for this wonderful explainer video.👍

With “no-gain” approach, Rode could have launched a game-changing sound interface. Something in a CloudLifter form factor, which would have just just a single microphone preamp, single XLR-in and single USB-C-out. And a little 48V switch. You just insert this interface (itself, being in a body of oversized XLR plug) into the back of any XLR microphone, and your XLR microphone becomes an USB one, with 32bit capture.

im back to 1990s 16/44 audio thank you very much

This is amazing! 🔥🔥🔥🔥

I think it's a good thing to have that option to fix things in post, but this doesn't change the recording process for me. I'm still going to adjust the gain on the proper level, since I'm on the job, I might as well do the job. It's just good to have that option later, if something unexpected happens.

It seems to me that indeed, 32 bit float does not magically remove any analog noise, but it does encourage designers to minimise that noise in the first place - it throws the focus back on to the very earliest points in the recording chain. So 32 bit float recording systems use multiple preamps and A/Ds, each optimised for a certain range of input level. That helps. And particular care will be taken over the design and implementation of any analog components before the actual preamps. A certain company whose name begins with "Z" has recently launched a one-piece stereo mic and integrated 32 bit float recorder. This seems to me to be an excellent approach because every part of the recording system, from the capsule to the SD card interface, can be optimised to integrate with every other part to the highest degree within the price point set of the device. The device I'm thinking of is actually pretty cheap, but for the money it sounds pretty damn good to me, because the designers have had to live up to the expectations that 32 bit float devices encourage. (Shame that a manufacturing issue caused it to immaculately record RF radio station breakthough too and they've had to recall it for the present... but once that's fixed, it could be a game changer in some sectors of the recording world). I look forward to RØDE producing something similar, but more upmarket, in due course!

How do you handle proper monitoring of a performance recording using these methods?

So... is there capacity / plan to upgrade the CasterPro II to 32 bit float then? When? Also, it seems you have not been responding diligently to an update which has made the mix minus function unpredictable and unreliable. When will you fix this?

I was hoping for a true objective explainer video, that will also mention downsides of 32-bit float. Like more work in post-production and file size. Completely failing to mention difference between gain staging a preamp and a digital file in DAW is also a bit disappointing. Hiss doesnt come from 24-bit, it comes from the preamp. 32-bit float is only game changing when you are working in a 100% digital world and thats okay. Its great for film makers for sure, because it allows for mistakes. But music studios will be recording in 24-bit for many years to come, even though our DAWs may be working in 32-bit world. I think most engineers understand the benefits, its just that dynamic range of 24-bits is absolutely enough for almost everything. And human ears like the sound of preamps driven to the edge, so we will always touch the gain there anyways. 32-bit is not superior, its just different and yes, its hard to understand because it cannot really be compared to video or images. But its a nice marketing video :)

Can you say please what the preamp type for each A/D converter is? Are they class A or A/B?

very nice choice of DAW to explain the video. I love reaper's workflow

Using 32bit since 1O Years on Stagetec Consoles . Nothing new in professional circles , but realy good also for consumers because thats where the lack of knowledge leads to clipped or to quiet Audio. Especially for Camera Guys ,..... those need it the most.

There’s still a quantization amount even in that adc configuration though. What is it?

Are there any (portable) audio intrfaces for the homestudio that can record at 32-bit float?

Nice bit about 32-bit recording, now do you have any products that support this? I've been waiting for a 32-bit float capable Rodecaster Pro.

Very interesting!

If there is a bit of hiss noise from a fen nearby the microphone at the momet of recording, turning up the gain should make it disappear?

how long will we see the double-precision floating point?

Great video.

Is a Rode Wireless Go III with 32 bit float in the works?

So what recorder do you need to do 32 bit float?

..almost forgetting an important point, normally a Daw uses an ASIO driver (for low latency in recording) with ONE audio interface, not multiclient, but in this situation this mic haven't an headphones output for monitoring while tracking, so I suppose it has a Rode Asio multiclient driver wich get possible to set the output of an audio interface while using Rode NT1 5th gen as input interface, right ? No issues with Daw compatibility and low latency at 48kHz with a medium cpu and soundcard ? Thanks again, have a nice day !

Hey how to record with backing tracks(usb mode), as this mic doesn't have headphone jack.??🤔 Does it have any direct monitoring system somehow??? Thnx 😃

I think the stacked converters make this more like light stops in visual, therefore HDR not raw.

Any chance the 32 bit float option will ever work on the Go 2?

What is the equipamento used in 3:06 ???

So my road caster pro is recording in 32bit? Or only new equipment still to come?

So there is a digital and a analog output?

As I'm creating my homestudio for my voiceover work (so not at all a sound ingenier or anything approaching), my question would be : what kind of audio interface we will need to have benefit from 32 bit float mic ?

What happens if I feed my interface a 220V signal, will it clip then?

If I record in 32 bit format, then in the daw after the recording when I need to increase the volume in the low volume parts, wouldn't the noise also increase? The bit format is related to the process of anolog to digital conversión. While highest is the number of bits you are using more precise is the conversión. I dont understand what is the relation between the bit depth and the dynamic range

When are you guys making a rode wireless go III with 32bit i would buy that.

Since I have been using zoom f6 last 3 years, and i go for 32 bit float everytime now and then . it's has become a new standard for me now. Press record and sleep, lol. Now in want this mic.

Any recommendations for interfaces that could record 32 bit? I couldn't find any yet. Only one I heard could do this was MixPre from Sound Devices

Thank you so much!!! I was just looking for how to professionally record my JawHarp, it sounds so good, like saw bass with vibrato effects, this way i can play it live with beats coming from my LaunchPad, i will have pre recorded pieces and intros will be live👍🤝

So if I record in 32 and capture all kinds of clipping, I can somehow make the clipping disappear?

But it doesn't solve electronic noise, or does it? So if somebody makes very quiet noises you still have "bleed" from electromagnetic noise sources, no?

This would be more useful if Røde could make a capsule that can take eq well, or one with a somewhat flat frequency response. Besides, 32-bit float recording isn’t very useful if you take the time to gain-stage properly.

So, if you are recording and you cant hear yourself because the gain is too low, just record it anyway and fix it later? Turn down everything else and turn up the monitoring output? am I missing something?

I wonder if cinema cameras will one day record in a series of base ISO's simultaneously, so the raw footage has a dynamic range of something like 30 stops.

I thought that the higher the bit depth, the more exact the sampled instantaneous voltage will be. Am I wrong?

Does your camera have to have 32 bit float capability for it to be recorded in camera? Seems like you will always have to sync in post.

One thing I don't understand is what the step range is defined as? Sure 32bit can do 1500db of whatever but wouldn't it be more useful to have maybe 200dB but be able to have steps of 0.0000001 db instead of like 0.001db?

What's the ENOB?

I've had noisy recordings in 32 bits when recorded with little gain. The noise floor seems to be a problem still.

isnt the noise floor of the electronics constant (if i would use a 32bit float audio interface)? so if i boost the audio in post i also increase that noise floor anyway? isnt 24bit enought with 144dB because the mic can handle noise till 120db anyway?? do you know an audio interface with 32 bit float and a noise floor similar to the mic?

Awesome review. I'm however wondering how you would use the USB in a 32bit recording in pro tools for example when overdubbing. I mean how would you handle the monitoring roles for the headphones for example? Thanks

i dont get how compressing the 96bit int output of the 4 adcs to 32bit float has anything to do with raw capture

The performance of this microphone has much more to do with the fact that it uses four preamps and four ADCs in parallel and then switches between them for the final output, rather than the fact that the final output happens to be in 32-bit floating point.

awesome. this podcast should be called "Rode Head"

I have a follow up questions then. How is it that there aren't any interfaces already using this? (Only Steinberg seems to offer 32 bit but it is integer instead of float.) I have seen plenty of handheld recorders now. But actual professional interface are all still 24 bit. Is there a reason why?

6:05 is that a power cord in an XLR input

Don’t you still need to set good levels in the analog domain?

does it work with any DAW?

How do I update my my Rode NT1 for this mode😭

Do Rode make any 32 bit float devices or mics? It'd be amazing to have that capacity built in to a self powered mic.

Will 32bit come to wireless go mics?

It's not that "not needing to set the gain to get a good recording" confuses me. It's all the other things I need to interface with and distribute audio to during a recording. Much more, how would this apply to live sound? Does the entire studio need to be "32 bit compatible"?

Something to bear in mind with floating point addition is that it is not associative, that is, a+(b+c) needn't equal (a+b)+c. In audio signal terms: if I have two signals, X and Y, and phase invert X to get -X, then if I first add the samples from X and -X before adding Y, then the sum becomes 0+Y=Y. If, on the other hand, I add X to Y first, and then add -X, I don't get Y, but rather a bit-reduced version of Y. (I'm not sure how relevant this actual is to audio production, mind you, just something that comes to mind when I think about how float's are represented in binary -- if all the signals being summed are substantially different, then this is unlikely to matter that much.)

I remember when we got Cubase vst32 in the late 90s. Game changer.

Why do I feel like the commentator is talking to me like I am a googoogaga baby

Who makes a 32 bit digital front end though?

So what you are saying is that 32bit float recording eliminates noise? I get that it would eliminate clipping, but the noise is the biggest issue I’ve found in recording. Mic noise, preamp noise, noise of interference, noise of power/bad grounding, etc. there’s no way to fully eliminate noise in my experience. I’m curious how 32 would do this. It would eliminate clipping yes, but i doubt noise.

How is the sound quality of 32bit float compared to 24bit fixed? I dont understand the technology yet. It doesnt have noise floor? I dont care about avoiding clipping but i care about sound quality

Great work, guys! I have a question. I want to recommend the Rode NT1 5th Gen to a friend of mine who is constantly clipping on Discord due to him using a Blue Yeti. Would this microphone help solve that? Or is my friend just a lost cause? 🤣

Do you need a special microphone to get capture 32-bit float ? Seems like a logical question that was totally ignored.

People still don't understand the difference between 32 bit integer and 32 bit float. 2 totally different things.

I will wait till Audio Science Review gives their verdict on this.

What's the dynamic range of this then? We talking thousands?

You can do the same thing with integers as well. Cameras use integers and NOT float when recording RAW footage. Seems like the marketing team is running this presentation who and want to jump the 32bit float hype train. A better analogy would be that you're recording the audio with multiple microphones from the exact same position with different gain settings. Then you can choose the recording with the best signal to noise ration after.

Finally. I've been wondering why that hadn't happened yet

quick question. if 1dB is the lower threshold of hearing and 130dB is the upper threshold of pain, why would anyone need more than 144dB of dynamic range?

how big is this file ?

Cool, so make a RODE Go III?

The interface has to have this for this to work. You can 100 percent clip every pre amp in every interface ever made.

we are believing here that the win stage is not important. For me it is important to have control of my gain knob. Decisions for a good recording start with the microphone and the preamplifier.

Is there any possibility that the next-generation audio interfaces will record in 32-bit float? That way, people can use all their mics in 32-bit, not just an NT-1 5th-gen.

I think the comparison with RAW Video is not correct. Even when shooting RAW you have to heavily adjust your camera: ISO, aperture, shutter speed - and that's because the sensor has a limited dynamic range. But I propose the microphone also has, even while it's analogue gear.

If they can do this with the next version of Wireless Go's that would be sweeeeeeeet

So as long as you have a strong computer and plenty of hard drive space clipping is a thing of the past👍🏽😁

I'm curious how you get 32-bit float recording from multiple microphones. Having e.g. 10 of these all connected via USB is surely going to cause bus contention, and if you connect by XLR, I presume you don't get the 32-bit float recording. I'm hoping patents don't get in the way of interfaces in general adopting 32-bit recording on their analogue inputs.

If this is true. We could eliminate the need for preamps. Hypothetically we could plug a mic directly to the ‘input’ of the interface and then adjust accordingly.

I'd like a wireless go that has this please.

So this is more like HDR photos where you take the picture at multiple exposure levels and later smash the range back down to something the display/printer can show.

no matter what. I cannot just not use UAD LA610 Mk II Pre for Gain Staging. Those ceramic hardware preamps are one of the best possible things that can happen with an input signal.

What was the point of the studio?

Hmmm IMHO some glitches: - DAWs today use 64 bit float rather than 32 bit float. - when captured in float format, even in the DAW (as using Floating Point for processing) clipping is not possible.Hence it in fact is not strictly necessary to bother about gain at all. While of course a kind of gain staging makes sense to improve optical representation of the signals, and to improve the working of certain dynamic and nonlinear effects.

Well, the analog noise will always be finite number. Not even speaking about ambient (acoustic) noise....

6:05 what did you just plug it to?!