We're talking about sound editors: not a gullible audiophile consumer pool.
Sound Editors or sound/music engineers can be as gullible as anyone else and there are several examples where they got suckered en-mass. The difference between editors/engineers and audiophiles is that they are more educated in the basic facts and the practical usage/application of the equipment and therefore many of the basic things like audiophile cable claims/marketing, etc., simply won’t work. To work on engineers, the concept of the new idea/feature has to be quite advanced (EG. Be based on areas of technology/theory in which engineers only require a moderate level of understanding) and the marketing has to be more sophisticated and include more objective details/facts. Furthermore, even if it meets the above criteria and is just complete BS, unlike the typical approach in the audiophile world, engineers will do objective tests, set up blind testing, etc., and it will eventually be out’ed, although in some cases this may take several years to filter down to all the masses. Therefore:
Clearly it's not some theoretical mumbo jumbo that editors who use it have no clue.
How is it clear it’s not some theoretical mumbo jumbo? And Editors do not have “no clue” they usually have a lot of “clue” but do they have enough in this specific instance to tell the mumbo jumbo from the facts? Apparently not, at least not the one you posted, who demonstrated some fairly rudimentary misunderstandings.
You might not be aware since it's new and doesn't affect your applications in a controlled studio environment.
It is new and I haven’t seen one yet but it does affect my applications, I’ve worked in TV/Film sound for over 25 years. And, a controlled studio environment is where it would very likely make a difference, because we have a very much lower acoustic noise floor than in pretty much any shooting location.
And again, we're talking about the file bit depth (let’s stop confusing it with processor).
I’m not confusing it with the processor, I’m explaining that pretty much any process carried out on an audio file, even just lowering the gain, yields a result greater than the bit depth of the audio file. So if we have say a 16bit file and lower the fader gain by say 12dB (12.04dB if we want to be precise with the math), the output of the channel is an 18bit file (unless we’re working in a 16bit mix environment, in which case we’ve still got a 16bit file because we’ve truncated the two LSBs). In practice the output would be a 64bit format file (assuming a 64bit mix environment) containing 18bits of audio data and a bunch of padded zeroes (assuming no processing other than just the gain change).
And sorry, having that I have worked with 32bit float image files for years on end, I know what the implications mean when it comes to representing the math.
And I’m sorry, having worked with cars for years on end, I therefore know how helicopters work. There are in fact quite a few similarities between the two and some of the knowledge from cars is transferable to helicopters but there’s also obviously some very significant differences and the assertion is nonsense/fallacious. In this analogy you are assuming your knowledge of cars is entirely transferable. I don’t know enough about digital photography/image processing to know how valid your assumption really is, but from what you’ve stated it seems to be “not much”.
You probably wrote all this before my last edit:
Correct but I’ll address it now:
Edit, when I look up 32bit audio files, I get the consensus that it's for new devices that don't need gain control (vs traditional recording technology where you do set levels to set the recorded range).
https://www.pro-tools-expert.com/pr...t-float-audio-when-is-the-best-time-to-use-it
Two issues: Firstly, “
the consensus you get” - Now we’re effectively talking about what “you get”, which is confused not only by your insistence to correlate what “you get” with what you know of digital imagining but also by your misunderstanding/misinterpretation of what you’re reading because it’s written for other professionals and therefore with certain assumptions. 32bit float audio files are not remotely new, as a writable audio format it’s been around for about 20 years and as a virtual/processing format for about 30 years. ProTools has had the option to record/write 32bit float files since v10 (released in 2011), we can take for granted that most engineers would know this.
What’s new is the claim that recorders/ADCs can actually record 32bit float and that it has such a large dynamic range you don’t need any gain. This is largely BS marketing but it contains some truth (and wouldn’t stand a chance of working with professionals if it didn’t) and is quite sophisticated. It therefore requires careful interpretation of the claim, for example, what is really the claim vs what is the implied claim vs what is it really doing? The first sentence of this paragraph was the implied claim!
To start with, what is meant by “record”? Does it mean an ADC that can output a 24bit (or 32bit) format file or does it mean an ADC that can actually capture 24bits worth of audio? If I take the output from an 8bit ADC chip, write it to a 128bit audio file and output that 128bit audio file from the ADC, is it a 128bit ADC or an 8bit ADC outputting a 128bit format file? The answer is the former, it’s a 128bit ADC, at least according to how we describe the bit depth of ADCs/DACs, even though it’s only really an 8bit ADC and the other 120bits are just random noise. In practice no ADC can achieve more than about 20bit. However, these new “32bit”recorders (there are 3 brands released in the last year or two to my knowledge) are quite a bit more sophisticated than just effectively lying/misleading about how many bits are actually being captured. Let’s say you have 2 different ADCs both working with the same input (mic output), but one has relatively high gain, the other a lower gain and you lower it’s gain further still in the digital domain. Then, you somehow mix/splice these two ADC outputs and output that “mix” to a 32bit float file. (BTW, none of the manufacturers explain exactly how or what they’re doing with this last step). Now what have we got? Well in theory, you could actually have 24bits of dynamic range (instead of just 20bit) or potentially even more, depending on how much you digitally attenuate the signal of the low gain ADC. However, there’s a few big “buts” here:
A. We’re effectively bypassing the analogue constraints of the ADC (20bit or so) by attenuating in the digital domain but it obviously doesn’t bypass those constraints in a DAC, any dynamic range extending below about 20bits will be lost in the DAC’s noise floor (not to mention the noise floor of the amp and speakers) and even if it wasn’t it would be way below audibility anyway. It also will not affect the dynamic range of the mic, in the case of say the CMIT 5, that’s 81dB dynamic (about 13bit equivalent) regardless of how much dynamic range the 2 ADC trickery allows. And of course, the mic can only capture 81dB dynamic range if those conditions actually exist in the acoustic environment. People speaking at say 60dBSPL in a room with a noise floor of say 40dBSPL is only a 20dB dynamic range to start with and nothing is going to change that, regardless of how much more dynamic range the mic or recorder has.
B. Clearly it’s not the case there is no gain/attenuation. There doesn’t need to be any gain controls for the recordist to set though, the gain/attenuation would be all internal and so widely spread that pretty much any mic (with any peak threshold, sensitivity or signal strength) could just be plugged in.
Regarding the new article you posted: It’s certainly a lot better informed than the previous one but there are still some significant problems. The examples given are fallacious/wrong (almost certainly inadvertently though):
“
However, if someone on stage makes a change either to their equipment settings or playing styles then it’s perfectly possible to end up with clipping on the recording. 32bit floating point audio allows the person recording the event to set gains with caution in mind, knowing that any clipping on the recording can be removed in the DAW …” - Hang on, how can the recordist “set gains with caution in mind” when you’ve just stated in big caps there is “NO GAIN CONTROL”? With 24bit recording and gain control there’s no reason you couldn’t allow 25dB of head room in this situation, I’ve commonly seen guitarists or others change their settings by 4-5dB, on rare occasions by 12dB, never come across more than 20dB though. That doesn’t mean it’s impossible of course, maybe there could be a situation where you might need 40dB of headroom but it would be exceptionally rare.
“
There’s a lot of fast moving parts and the unexpected can happen at any moment. In some cases the scene can be cowboys whispering threats then gunfire!” - How would that be unexpected, don’t they have a script, are they just making the film up as they go along? In most countries there are very strict laws about when, how and by whom a firearm can be handled and discharged on set, so unexpected gunfire would be illegal. In the USA the laws of firearms use on set are not quite as clear most other countries but we’ll probably get more clarity after Alec Baldwin’s manslaughter trail! So, this scenario would be expected and we could deal with it with existing 24bit technology, in fact this sort of scenario was well handled when we just used 16bit recorders.
“
imagine capturing sound as a jet airliner comes into land. The shot starts with the plane half a mile away and ends with the rubber hitting the tarmac. At the start of the shot the sound of the jet is hardly audible but as it hits the tarmac the sound is deafening.” - Deafening indeed, probably 150dBSPL or more if you’re fairly close, what are we going to record this with? We can’t use our lovely CMIT 5, it starts going into audible overload distortion around 130dBSPL. Some dynamic mics will handle 150dBSPL peak without much distortion but then it’s not going to pick up the sound of the jet half a mile away. Our limitation here is the mic, not our recording bit depth.
“
Having researched for this article, many location engineers were sceptical of the 32-bit float theory to start with, but having created a 32-bit float workflow, said they wouldn’t go back.” - Now that I wouldn’t argue with! It takes time to plan what gain you’re going to need and then setting the right gain for each mic for each shot. If there’s no hit on audio quality what location engineer wouldn’t want to simply plug in their mics and never have to think about or set gain? This 32bit location sound workflow doesn’t avoid the issue of gain staging though, it just kicks the can down the road, as the article correctly asserts, “
In effect it passes the problem of gain staging onto the post stage in the DAW.”
Using 24bit recorders, there’s no reason why there should ever be any digital overload clipping, however I have occasionally seen it and pretty much whenever I have, it’s been user error (due to a film maker trying to cut corners and hiring an inexperienced recordist) but there are potentially very rare conditions where even an experienced recordist might be caught out. This is where 32bit recording could be useful because as I mentioned before, it would allow recovery of digital clipping (where 24bit doesn’t). Although it would somewhat open the door to cheaper, less skilled/experienced location sound recordists if they don’t need to know anything about or correctly apply gain-staging.
G
. Displays are now capable of displaying up to 10bits, so there is UHD content that's either 10bit or 12bit (that gets adjusted to 10bit on the fly with embedded metadata). When we were talking about HDR for photography, it has referred to getting a file that has a dynamic range that exceeds the camera sensor (bracketed exposures saved to 32bit). It's now ubiquitous with consumers that learn about the difference between video that is SDR (8bit) and anything above that being HDR. For many years, cameras have been able to record 12 stops of light. And professional video cameras can record 16bit RAW: so one of the tasks an editor does is grade the color for display media.
