I see you're really trying to understand this stuff, so here's some extra info for you. There's a limit to dynamic range at several places in the recording chain. Starting with the mic, the maximum SPL level is a physical problem and exceeding will result in physical damage. The microphone will output a voltage level in a certain range. This is then (usually) fed into a gain amplifier. This amplifier is used to keep your maximum SPL in the voltage range of the next step, which is the Analog to Digital converter (ADC). The gain amplifier has an electrical limitation on it's maximum voltage as well, and passing it will cause clipping or saturation depending on it's construction. Next the ADC takes the voltage provided by the gain amplifier and converts it into a digital representation. This is designed to convert a specific voltage range, and only that voltage range. This is an electrical engineering thing, and there's basically no way to measure any possible voltage. This range is designed to be good for a unity gain amp stage, and designed with the electronic noise present in the previous stages and it's own circuitry. Typically, on the best audio IDs, this will give a signal to noise ratio of 120db. Exceeding this range will turn on clipping lights or, in the case of Zoom, max SPL exceeded warnings. If you're exceeding the ADC max voltage, you're expected to turn down the gain amplifier or otherwise attenuate the signal. The gain is also used to put very quiet sources more fully into that range, though less so. As the (pulse code modulated) ADC is designed to measure over a specific value range, it is able to convert the ratio of voltage/(voltage max) into a corresponding digital range. Here's an important point. ADCs are signed integer based, at a fixed internal bit depth, i.e. 32 bit, due to how we measure voltages in electronics. This is basically a ladder of resistors that the voltage climbs, each step indicating a certain minimum voltage level. They will produce an integer over the range of -(2^bit_depth) - 1 to 2^bit_depth. If you want less bits, or a different format, than the internal representation the result is dithered and converted into the output type requested. The also tend to sample at several times greater frequency than requested and resample the output for reduced aliasing. This is the first point where floating point can appear but it's not the fundamental measurement being taken, and this is important. This means that floating point is not fixing anything about gain staging up front, and it's not better than the original integer measurement. It may even be inferior for storage from a strict perspective as floating point is actually quite bad at representing some numbers. You can read all about the weirdness that is IEEE 754 if you want, but it's not that relevant here. After the ADC is done it transfers the PCM waveform to a computer, file, or what have you. In the case of handheld recorders, it's going straight to file. In your case a 32bit float WAV file. You take your file, you put it in the DAW and then you process. Here's where a bit of history comes in. Computers used to be pretty bad at floating point. They used to be pretty bad at 24bit integer math too. Even switching a Pentium between integer ad floating point registers (the CPUs internal working space) was a bad performance hit. Early computer sound processing did all it could to avoid that. Most things didn't default to floating point processing, and didn't even default to 24bit processing. 16bit integer was pretty standard on computers, and digital hardware was often using 8bit and even 4bit far more than you'd expect. This meant that even mixing down tracks was so ridiculously headroom starved that anything other than meticulous gain staging would result in huge noise floors. Thankfully times have changed, processors have changed, audio engineering has moved on. 24bit integer is the default capture format because it's compact, fast to process, and covers all the realistic SPLs we'll ever get past an ADC. Additionally DAWs *all* universally process audio as 32bit floating point after they load a file regardless of what the file format is. You can probably even find a setting to process your audio as 64bit double precision floating point if you'd like. But even loading an 8bit wave file, the DAW will convert everything internally to 32bit floating point. This means loading a 24bit integer wav file is equivalent to loading a 32bit floating point one at the cost of space, and possibly compatibility. The history of computer audio means that some things are more compatible and universal than others. Floating point processing, having come in relatively late, being not exactly native to PCM, and taking 4 bytes to store instead of 3, means that some waveform editors (The standalone kind like audacity) may not support 32bit floating point WAV files. Most will, as will most DAWs, but the testing on them is less thorough as they're used less overall. Standalone versions of plugins may also struggle here, as this is just not the expected format for WAV files. The same is true of VST plugins supporting high sample rates or double precision floats. Good ones will support, most made with good libraries will work, but some may have odd problems. Why take the risk? Floating point processing is and was a major coup for computer audio processing, but it's not needed or even advantageous in recording or storage. Set your mics gains' so they don't unintentionally clip, and you're fine. A side note as a programmer: In computers, integers aren't the same as fixed point or floating point numbers. Integer numbers have perfect precision for the set of integer numbers (the kind without fractional components). Fixed and floating point numbers both can represent the set of real numbers with imperfect precision.
Jesse, thank you so much for taking the time to go into such detail. I’ve read through this a couple of times now and thinking about the implications. Thanks for clearing up my conflation of integers and fixed point. Have I fallen for a bunch of marketing crap? It sounds like there are tradeoffs on the recording side (particularly the further a value is from 0 whether negative or positive?), but is the problem mainly on the processing side, after bringing the file into the DAW? At the core, what is the impact on the sound, if it’s even possible to put that into words?
@@simonpeck So I looked more into what Zoom specifically is doing as I wasn't aware they were going so hard on 32bit float. They're doing some nice engineering things, but also being a smidge overzealous with the marketing. Zoom is making their things with _two_ ADCs. They set one to high gain and the other to low gain. They got rid of the adjustable gain amp, instead tuning the gain of the ADCs for the max SPL of the microphone. Gain controls it gives you are digital amplification post sampling. The ADCs are still regular ADCs, but the signal is processed by a microcontroller to produce the final measurement. In float recording mode, they use the signal from both to determine if something is above or below 1.0. In integer recording mode they _still_ use both ADCs but they use the best ADC measurement for the signal at the time to keep it from clipping. They don't say exactly how they're combined, but this is my guess. They could also be doing the same thing as some oscilloscopes and splitting the input range across both of them. That would give you incredible dynamic range at the expense of lower normal signals. That's not really a problem with SPLs of 140 or less though, which still fit into the 24bit space. The design reduces the noise floor so this all works out well. This is pretty slick, and a good example of a well tuned system from an industrial design perspective. It's not something you couldn't do yourself by recording a signal at two gains, and it still has compatibility problems, but it's neat and well packaged. The magic is really in the dual ADCs though, not so much in the 32bit float. Things to still note are that even in float mode you _can_ still clip the ADCs. It's just much harder. You also get the same full dynamic range in 24bit mode. If you're recording to files, the 32bit float WAVs are still a possible compatibility problem, but you''d only need to normalize and convert them with something to use them anywhere. If you're using it as an audio interface, then you may have issues recording directly in 32bit float. Zoom has a compatibility table, and I'll point out that it's probably also only valid with ASIO drivers in Windows. WASAPI does have 32bit float format support, but it's pretty unused and thus not tested well. You might end up with not being able to record it, or recording it but any peaks past 1.0 are clipped, or other strange bugs. I would test anything before using it for anything important. As you get the same not-have-to-set-or-worry-about-gain in 24bit, anything that doesn't behave well has a good fallback.
The noise of your analog input limits the dynamic range to about 110db/20 bit at best. 24 bit recording with levels set in the middle, with plenty of headroom should be good enough then. The 24 bit master can then be used to create a 16 bit file for listening... still with a 90db dynamic range (which is huge for popular music).
There's no question that 24-bit is enough. The purpose of this techniques is more about how recording in 32-bit float can change and simplify the workflow, especially when it comes to recording with a dozen microphones. It isn't an argument that the dynamic range or sound quality of the final product is better.
Interesting video! Was hearing a bit about 32-but float about 10 years ago and people seemingly stopped talking about it. Happy to see such a good application for it with a straightforward explanation.
I was just wondering, how can this channel only have 400 subscribers?? That will be over soon, because this is exactly what is good content about 32 float
I appreciate the segmentation, straightforwardness, and overall flow of the video. Your thoughts and opinions feel very grounded to reality and you support those with solid information that you've acquired through experience and ultimately by just being familiar with the equipment in question. Keep up the great work, you've most certainly earned yourself a subscriber.
5:34 the converters aren't actually 32-bit float, they are 24- or 32-bit fixed. The floating point format is something completely digital and can't be recreated in the physical world. They instead use two, or three, converters at different gain levels and the 32-bit float files are then calculated from the combined signal of the converters. This is why you can always recreate the waveform even if one of the converters clip, you just replace the clipped part with the signal from a lower gain converter (although it's a bit more involved than that).
Yes, but that is an implementation detail of the Zoom F6 (and other 32-bit float recorders currently on the market). For practical purposes, it doesn’t matter. As a recording engineer, it doesn’t affect the approach to recording.
Thanks for that hint, that explains to me why the whole thing is working in the first place. From that information it is clear that creating gear that covers a larger range is not impossible but will be exponentially more expensive because for every increase in dynamics you need to add: A) another gain stage + ADC B) more processing power C) more internal bandwidth D) possibly another storage device because at some point your SD card will bottleneck
Also, there is not only the upper limit but a lower limit as well, which is mostly the preamp in the Microphone, as well as thermal noise in cables, resistors and so forth, which at room temperature and a bandwitdth of 20kHz is roughly -130dB. There's also something like bit flicker and clock noise in the converter, so the real dynamic range of any device is much less than what's theoretically possible with 32bit float, even if you split the signal across multiple amplifiers. Also, unless you are in a well treated acoustic room, the normal room noise kicks in much earlier than the thermal noise. At >60dB of amplification, it usually becomes quite audible. Which is to say that you will never get rid of the task of getting a decent signal to noise ratio in the first place. But then again, drums are a very loud instrument, so you get a good signal to noise ratio with almost any recording setup (unless you play very dynamical with extremely quiet drumming mixed with normal/loud drumming). With drums, any modern AD-converter should give you a good signal to work with. In my experience, the most important factors are: Mic placement, mic choice, acoustic sound of the drums, acoustic sound of the room... and the rest is absolutely minor compared to these four things.
@@simonpeck N J has a valid point. On audio interfaces that use only one converter per input channel the converter is always the bottleneck for the signal. It can be 24 bit or 32 bit but it is always fixed-point, so you lose any advantage that floating-point offers already at that stage. Zoom tries to overcome this bottleneck by using two converters in the F6, each with its specific dynamic range. Your example with vocal and drums is basically what Zoom describes in their manual. So the first answer to the question raised in your video title: Not everybody records drums like this because not everybody has a multiple-converter interface. Your thinking of recording "dry" and do anything else later in the box was already envisioned by some producers in the 1980's. I saw a documentary about Trevor Horn's production techniques. They tried to do "in the box" whatever was possible to do in the box back then. That leads to the other answer to the question: Plenty people working in music studios still behave like in the analogue days.
@klauba Yes, you’re right about the bottleneck. You must have a 32FP recorder or interface to do this. My question was how come “nobody” records drums like this, not “everybody.” I was sure others must, but I hadn’t seen any other evidence of that online. I was hoping those who do might be drawn into the conversation, and a couple of people have. 🙂
I have been recording in many cases where all I had was an interface and 8 channels ADAT with the equipment in the same room as the instruments being played, Thus not being able to make and compression or EQ decisions, They would be left out of the signal chain altogether. Sensibly I would set gain levels to be safely below clipping and proceed to do up to 1- 1/2 live takes, All the decisions had to be made later. I found 24 bit was perfectly fine to get good clean captures , even from substantially boosting quieter parts. What a great time to be making music in! and great vid!
Thanks for making this video, there have been many times where I have had the opportunity to record in 32bit float but chose not to due to my gear “not being professional” by some opinions. You have inspired me to try it out and work with the post record workflow you described. I’d love to see your remote setup in a future video
Yes! Try it out. What type of audio are you recording? There does seem to be some trepidation around the topic, but don't let that stop you. It's freeing to know you don't have to worry about levels. And in the case of a drum kit, it's easy to make a mistake when setting the gain correctly for 12 microphones.
@@simonpeck I mostly work on recording my personal music ideas. Also, I am starting to get a little more serious as a sound designer. Finding sounds with a field recorder and not “needing” to adjust the gain constantly will be a timesaver.
floating vs fixed as a choice for digital audio is somewhat analogous to the use of dolby in analogue audio, in that it is advantageous to s/n ratio when used properly, but goes horribly wrong when it's not; bit-rounding that can happen to digital signals if mishandled will give unpredictable results because vital parts of the number are lost. this doesn't happen with fixed point, as you say.
Amazing, very informative video thank you for the post. I've been recording in 16 to 24-bit wav audio over the last 25 years. I have never had enough money to buy a 32-bit float system. Recently using a rode podcaster i used the the 32-bit float option. Like you mentioned, I was able to change all of the levels after in post from tiny little laughter's too loud comments during the podcast, game-changer for sure.
making electronic music for the past 15 years I never thought about using 32 bit recording/files (mainly because any genre of electronic music is rooted in old, far less capable samplers than the machines we have today), however, this was such an amazing and interesting conversation on the whole topic!
Thank you all for watching this video and your incredible, detailed, well-articulated comments. I'm blown away by the response! CORRECTION: In the video, I mistakenly said 32-bit float allows for 4.6*10^38 "states." What I should have said is that it can represent numbers within that range. There are still 2^32 states. Because some bits are used for an exponent, there are fewer available for the mantissa, meaning 0s are added, resulting in a larger distance between values at the extreme ends of the positive and negative values.
Nice subject. Recording is more than mp3, 16 bit, 24 bit or 32 bit. Musicianship, song, arrangement, production, instruments, acoustics, microphones, mic placement, pre amps, mixing, mastering, exporting. Audio/video. Light. Regarding recording technique: Noise of electronics, clipping, noise of drums, tuning, noise from outside, noise or degradation of signal in software or signal chain. noises in room acoustics, etc. etc. Phase correction can improve significant a recording. Noise really is irritating and can destroy a recording however sometimes a noise gate can help and adding some effects like subtractive eq sometimes also helps. There also are software type noise cleaners. Multi track is often more intensive than just mono or 2 mic stereo. Adding a lot of data types together like several audio tracks and several video tracks can take a lot of time for the alignment. Recording on pc/laptop does not require the sd card and sometimes it takes a lot of time for datatransfer and ...have a rough mix available at the time of recording (pc/laptop) may help to improve the artistic or technical performance during the recording session. On the other side: many people use recordings of iphone or smartphone and ...after mastering with iZotope that can sound OK for the purpose: showing you can drum or had a nice gig, a nice band or have made a nice song or just had fun.
@@HenkJanDrums Yes, this is all true. In my own recordings, I rarely, if ever use all 12 tracks. I'm always swapping mics, trying out different positions and sometimes mic the same drum two different ways. Later, I listen to how it sounds in the mix to decide what to use. Phase is often the reason to leave something out. But sometimes it just sounds better without certain mics in the mix. The target style for a particular song is another major factor.
Recording (in terms of converting from analog to digital) is always done in fixed integer (16/24/32 bit). You MAY choose your recording-file format to be 32 bit float, but the dynamic range in that format is also only 24 bit, because of the nature of the IEEE-754 floating point format. You will even get more dynamic if you record in 32bit fixed integer, because then you really have 32 bits of dynamic range, BUT, this is really over the top. I don't think, that someone can hear the difference of 32 bits vs 24 bits dynamic... You will even hardly find AD-DA converters doing real 32 bit conversions seriously, and very often the internal signal chain of plugins is also only 32 bit float (=24 bit dynamics), so, recording anything higher than 24 bits is practically more or less useless, most of the time. In 32 bit floating point, you have an effective dynamic range of 24 bits. Other than in integer formats, you can not clip the value range of float samples that easy, which maybe is, what you meant, BUT, always keep in mind, that even if you can produce values outside the range of -1.0 to 1.0 during mixing (and also when rendering to a file), there comes the time, where the data wants to be translated from digital to analog, again. And this is the moment of truth. Everything outside -1.0 / 1.0 will be clipped. Maybe some players have intelligent built-in limiters, maybe. So: If you like the "sound" of 32bit float much more, then you most likely like the "sound" of your limiters. To really have MORE dynamic, you should record in 32 bit fixed integer, and put your DAW into 64 bit floating point mode (if that is possible). THEN (and if all your plugins support 64bit audio processing) you really will have more dynamics during mixing. But whether your monitor / your ears will hear the difference, I don't know, but I would bet, that (especially in home-recording scenarious) you won't hear ANY difference.
You captured what I was going to say in more detail and clarity than I could have. Thanks for pointing out what I think is one of the key sets of facts.
Not in this case! The Zoom and sound devices recorders use dual 24-bit converters on each channel. One handles high gain and the other low to record in 32-bit float. When choosing this format, the limiters are bypassed.
@@simonpeck I have to admit, that I was not aware of that technology... And it indeed sounds cool, in terms of "setting proper recording levels, that do not clip". While many of the things I said are still correct, I was really wrong about the dynamic regarding capturing the audio. I am sorry for that. Obviously, the sampling process produces 2 different values per sample. One high gain and one low gain, then the high gain one clips earlier (depending on the gain differences), and still a valid resulting sample value can be reconstructed by using the low gain value, to reconstruct the clipped sample. And yes, in this case you can have more dynamic headroom, depending on the overlapping range of both gained converters, and I was not aware of that. But on the other hand, you really need to work on that recordings (as you showed in your video)... and for ONE thing it is great: you do not have to care too much, to get the right recording gain level. Which indeed is somehow "cool"...
My thought exactly - not only there is a noise floor on equipment, there is an “acoustic noise floor” too - the quietest studio will always have some kind of background noise.
Preamps and other segments of the audio chain up through and including the ADC are not necessarily linear. This approach MIGHT work well with awesome preamps, ADCs, mics, room sound management, etc., but even bleed will add pressure to the components in the audio chain and color the sound. And if you set the gain too low, you’ll have analog noise to deal with when you raise the levels of the channels you want louder in the recordings. In a perfect and perfectly digital world, this is a great way to skip over some of the work you’d normally have to do in setting the gain for recording drums. But, this is not the world we live in.
This was a very interesting video. I think some clarification of the pre-amp stage vs the AD conversion is important whenever talking about this advance in tech. The max input level of 131db is the point where the analog circuit before the conversion runs out of headroom. Unless a person is strictly discussing a conversion only unit, that limitation is going to be an issue with any system offering 32BF conversion. Then again, if you're working with converters that do not include pres, you're probably well beyond any discussion we could have here. Eventually, we may actually be able to record audio in a RAW format (like images and video). Exciting stuff. New sub. Thanks.
I've noticed that the same limitation exists across manufacturers. Can you shed some insight as to why that is? Why couldn't someone make a pre-amp stage that could handle, say... 150dB?
Dynamic range isn't the only consideration. Virtually all analog gear have a noise floor that is audible. Raising the source level relatively high above that noise floor ensures a better signal to noise ratio and a cleaner sound.
Generally a great video about an often overlooked topic. Here's a few things I'd like to point out tho: 1st) Regarding the maximum SPL, I don't see a reason why there should be a limit on anything that isn't a microphone capsule. All the microphone does is convert the atmospheric pressure waves (SPL) into a voltage. The higher the SPL, the higher the voltage output. This is heavily influenced by the sensitivity of the microphone. So, for a less sensitive ribbon microphone it would take a much louder sound (higher SPL) than for a sensitive condenser microphone to output a voltage that's higher than the mic pres can deal with. 2nd) As some commentators here have pointed out, making use of the entire theoretical dynamic range of 1528 dB is technically impossible. However, neither do those recorders simply store the data in a different format nor does a 32 bit converter chip even exist. What's really going on inside those units is a clever combination of fixed-gain mic pres and regular converters. Think of it like running your microphone signal trough a signal splitter and feeding it through multiple preamps with different gain settings in parallel. The outputs of the preamps then gets digitized through regular converters and are digitally stitched together. You're only using each mic pre at/around its ideal operating point. This way your gain setting will always be correct, which makes 32-bit recorders an excellent choice for any situation where levels are changing rapidly or are completely unpredictable. 3rd) I really like your idea of recording the drums as cleanly as possible and then adding any coloration later. But 32-bit won't really offer you any higher fidelity than a regular 24-bit recorder with properly gained preamps, as the fixed-gain mic pres still have a noise floor that won't be any lower relative to the operating point aka your signal. However, I can totally see how not having to deal with setting levels can make the whole operation a lot easier, especially when doing everything yourself. 👍 Here's a better explanation than I can deliver in text form: th-cam.com/video/NTuJ1fk3PsY/w-d-xo.html
Thanks for your comment! Yes, agreed with most of this. The only thing that I would point out is that most microphones have a higher max output than the max SPL of the recorder’s input circuitry of 131dB. Since you mention ribbon mics, one of the mics that overloads the F6 is an active ribbon mic on the kick head. I put a 10dB pad on it. Also, I’m not claiming higher fidelity, but rather an easier workflow with a recording that sounds good. I intentionally left out how this specific recorder works, because I wanted to talk about the principle of recording in the 32-bit float rather than focusing on any specific manufacturer’s implementation. However, since there seems to be a lot of confusion about it, I may make that video.
Great explanation. I use a sound devices mix pre 6 ii and 4 of the slate ml2’s. It is such a great combo to set up my drums in other places and record. Then bring into the big system.
Did not understand all the tech-babble (yet), but sure sounds like some innovative way to record/mix drums (I'm a drummer myself), ... so kudos to you and all the time you spent to come up with this technique and even made a vid for all of us to learn about it. A+ from me.
yes in your case you can save some money. thats a good reason to do this. otherwise there are other downsides involving the double-AD converter design and other stuff that is impossible for amateurs to discuss. you can read about that in bob katz mastering forum or in discussions including Lavry Engineering or daniel weiss, which are some of the best converter designers. btw stagetec has done this for years on a professional and very expensive level, but they didn't sell many consoles.
Nice mix of video production tips and audio production tips. I'm more interested in the audio stuff personally but always nice to know how to improve pictures. Cool channel 👍
For almost 15 years i record mp3 320 kbps up to wav 24 bit with zoom H4, H5, R24, and L20. Live mostly stereo (Zoom H5) and in studio multitrack (Zoom L20) up to more than 16 microphones nowadays. For stereo live recordings I use a modern version of Soundforge Audio Cleaning Lab (24 bit), that has a super fast workflow for making normalised and mastered songs out of a live recording. For home studio Zoom L20 multitrack recordings i Use Studio One professional (64 bit Float) . I record with recorders because they dont have fans like mixers or pc. I record with low gain values and do the processing in DAW afterwards. In my DAW I only work with normalised tracks so ...new recordings are mixed already very well when imported to older projects. With Zoom H4 and now H5 I always have a good live recording.The Zoom gear has almost no noise levels for serious hobby. However some cheap microphones may have noise levels. I can recommend iZotope for mixing (Neutron) and mastering (Ozone) it has automatic (analyse audio) settings that are really great. They can be set to drums but also to snare drum, bass drum, toms, percussion, acoustic music, detailed drums other styles of music, etc. The Zoom L20 has better pre amps than the R24. The whole process from recording to video did 10 years ago give a big loss in the video software. I now use modern Vegas Pro which has very good audio. To show how quality was many years ago and how it is now: see 2 video links of same song, same drumkit, same drummer but Zoom L20 in stead of Zoom R24, better mics, and most of all better video software. All recorded 24 bit ...than and nowadays. Modern mix th-cam.com/video/xe1couRF5Ms/w-d-xo.html Old mix 10 years ago th-cam.com/video/L1k9u7dVE9Q/w-d-xo.html v Modern drums th-cam.com/video/WmFeKEgfyWE/w-d-xo.html Modern drum time aligned (edited towards 100% exact timing) th-cam.com/video/GdOy_U_tSG0/w-d-xo.html Same type of soft acoustic drums 10 years ago. th-cam.com/video/uAs0w4DvO5A/w-d-xo.html
At 5:10, I don't understand the claim about different numbers of representable states for 32-bit integer vs float. With 32-bits it is possible to have at most 2^32 different states, no matter the data format. Integer format uses these states to represent equally spaced numbers, whereas floating point format has numbers that are closely spaced around zero and farther apart the larger they get.
It appears then that you are recording the drums "wild" outside your DAW. If so, for me this would make keeping takes organized a hassle as trying to do things like "punching in" a total nightmare. Plus all of those performances would need to be hand synced when laid into the DAW - meaning maintaining the same time placement of what was originally played a crapshoot. And all to preserve a dynamic range that no drum performance would ever need or take advantage of and avoid the hassle of setting levels - which as someone who has been recording drums in home studios for 30+ years, I find I only have to very infrequently do. Moving from full on rock playing to playing a jazz ballad with brushes would require a level reset - but staying within the range of most rock/pop/country/funk playing really rarely does. Just seems little (to no) advantage, yet more hassle compared to just having a standard 8 channel interface (or 2 of them) that stay with the drums - with the DAW session on an easily transportable laptop. But an interested thought...
I will say that not being able to punch in is probably the biggest drawback to this method. It isn't impossible to work around that, and because it isn't as clean as when working with a DAW, it does create a little pressure to record a better performance. So I wouldn't say it's all bad. However, aligning tracks is not an issue. These recorders can be configured to produce a multi-track, "poly" wav file, so all 6 tracks are in an individual file. If you use multiple recorders, there is at times one alignment procedure, but it only takes seconds to complete. In my next video, I'll show you how I work with these files and how I handle handle takes, since you asked. Since it sounds like you have rich experience in this area, I'd love any insights you have on that recording process. Thanks!
@@simonpeck To clarify, I wasn't speaking of keeping the drum track files lined up (that would be relatively easy with the recorder producing individual files - as they all start at the same time, which when recording a whole set, they would). No I was referring to laying them against the tracks, you are overdubbing to. The drum tracks as a group would be "wild" to the reference tracks - and thus the "feel" of the performance against those tracks as played would not be discernible - basically left to the ears of the person adding them to the DAW. Again maybe not the biggest deal - but personally, I like to know where I played things against the rest of the music, before I start moving things around - either by hand or by timing tools. Again I can certainly imagine the ways to make this work - I just personally couldn't imagine making those compromises for what seems to me to be so little advantage (if practically any at all) You asked "why doesn't more people do this?" - and the answer was sort of baked into all of those videos referred to - but then sort of dismissed. Things like not being able to punch, having to constantly manage, transfer and re-align track/files, not being able to easily listen to performances back while recording, having a process that really dictates going for complete takes... all of these extra inconveniences chew up time and energy... for a process that is already left brain/right brain demanding... where every added technical step just sucks away the player's ability to focus on playing and the music... Maybe soon there will be interfaces that will record into the DAW in 32 bit float - then we could have these advantages without suffering the workarounds. But until then, I can't see the answer to your original question being anything other than...."because it's not worth it". By the way - while I'm sort of busting your chops on this, I've always been enjoying a lot of the other videos on your channel. Good stuff. All the best, David
@@DavidCrigger Oh, I see... I do play against a source with in-ear monitors. I haven't done this, but you've just given me the idea. to try it out: I think that monitor source could be fed into one (or two) of the inputs on the audio recorder. The inputs can be configured to accept line level. That might solve part of the problem, though it isn't as elegant as a traditional approach. Incidentally, Zoom just released a 32-bit audio interface and Reaper supports 32-bit float recording, so I think we are on the cusp of having the best of both worlds. That was my hope and prediction at the outset of making this video. The Zoom only has 2 inputs, but maybe Universal Audio will add something to their lineup in the near future. I suppose one could get 6 UA232s and create an aggregate device for 12 inputs. 😂
Although there are good reasons to use 32bit float but you can't just ignore everything else. There is still a signal to noise ratio you want to watch and the dynamic range of the analog part of your chain prior to the converters can cause problems. In addition, there can be sweet spots where the actual equipment sounds the best.
Good vid, good process. I'd suggest, as a guy who used to record live jazz, that you try for a point where you use two mics max to record a kit. W/out deep explanation, one can get full flexibility at mix time with a clear, two mic recording and avoid phase issues, and even the need for processing beyond limiting and verb. The difficulty of "good drum sound" is reassembling a kit with 8-100 mics. If you can play somewhat evenly balanced, you can avoid a gaggle of mics, and multi-band compression can give you eveything you need at mix with a simple stereo mic pair at MS or XY config... In short, too many feel the need for too many mics. Most I've ever used on a kit is 4. Give it a go. The clarity is sweet.
Thanks, Roth! 🙂 I started with 3 mics, used the Glynn Johns method and slowly built it up over time. I regularly try different configurations from 1-18 mics and yes - phase issues can be a headache. I have loosely planned to do a series of videos using a different number of mics each time if that’s of interest.
@@simonpeck Second read of my blathering sounded a little *actually.* I should say "I spent thousands on drum mics only to use 2-4mics." Sorry if I came across like a net know it all.
All the mention of max decibel levels is great but it really doesn’t matter. What matters is dynamic range. If you’re recording a drum set, you’ll maybe have a what, a 20-30db dynamic range in even the most extreme case? In your configuration, you are still using analog mic pre’s and A/D converters, subject to S/N ratios and clipping. You can still normalize/boost a quiet part of a 24bit (or 16bit) recording. That’s been happening for 40 years. Maybe if you had a a mic that puts out a straight 32bit digital signal (I know Rode makes one) this would make sense, but I can’t think of many scenarios were I don’t want to take a few minutes to get good levels on the front end and do it later.
The point is not to celebrate any specific recording format - it’s to make the process faster, easier and less error prone, while focusing on the sound where it matters most: the drums and mics. I want to use my favorite mics for each surface. Then, good levels can be set in post, believe it or not. Do the drums sound bad? I probably should have had longer clip of the drums mixed, but I can post that in another video. For the past 40 years, all of the production hardware has had to be within a cable length of each other. In my setup, the drums are in a completely different physical location.
@@simonpeck you’re title says “how come nobody records drums like this” and I explained why. Yes, levels can be adjusted in post, wether 32 or 24 or 16 bit, that is sometimes called mixing. Not sure what you mean by in a different location. I’ve had various 16plus track mobile recording rig (usually 2-4 rack spaces now a small digital mixer) for going on 20 years now. For me, I’d much rather take a couple minutes run in the zone it runs optimally. Again, analog in means S/N ratio is a factor and clipping is still a possibility at both the mic pre and A/D converter.
@@zagatoalfa Got it. I would love to know more about your mobile recording rig! Yes, clipping is still a problem in my situation, which I talked about at 08:30 - I used a Shure input attenuator on a couple of mics in that situation. What I meant by "different location" is that my drum studio is not in the same location as my home studio. So it's basically a mobile recording setup. What do you currently use for yours today?
@@simonpeck for my portable rig in the way back I’d use a digi002/003, a Behringer 8 channel pre converter combo, and a Sytek MPX 4a for four “money” channels, giving me 16 mic ins. More recently a Behringer XR18 has sufficed for full band location recording. I’ve sold the XR18 to help pay for a Yamaha DM3 that will give me 16 mics into laptop via usb and be generally more useful for my present needs. I’ll note, all of these set ups do require a computer, though there are a number of companies that do multitrack stand alone mixer set ups these days. I did borrow a Zoom F8 in the early days of the pandemic to record rhythm tracks for a friends band in an alley. Battery power was a good option at the time! Another note about setting levels when tracking… if your levels are super low or all over the place, it’s going to make monitoring what your recording pretty difficult. If you’re recording more than one person, and they are on headphones, and the levels going to your DAW aren’t adequate, it’s going to be more trouble getting headphone mixes sorted than just tracking at a good level. Best way to have a bad day in the studio is to have a bad headphone set up.
I’ve been drumming for more than 20 years and a software professional for more than 14. I just bought recording equipment for the first time. I’m intrigued by this technique which reminds me of light field photography. I hope my equipment lets me capture in 32-bit float format.
I probably should have mentioned it in the video - what you’re probably hearing is the combined self-noise from 12 microphones, one of which is a tube and two of which are ribbons. The mic I normalized for voice was about 30” away from my head. The noise floor in the room is 28-30 dB. There is no way to set levels in 32 bit float mode.
Without Wordclock, I’m curious if you notice any issues with sync-ing the two recorders 6-track cards together? Looks like you’re just eyeballing it. I imagine they probably start out pretty close, but long takes will drift apart since Timecode is merely a Timestamp at the head of each file. If you split a test tone like a 1k sine wave source into both F6 recorders and let the take roll for a while, I bet you’d hear (and see) the drift when you line them up in the DAW.
Good question! The drift of the F6 is 1 frame every 48 hours, so if they were drifting in opposite directions that would be 1 frame in 24 hours. I haven’t found it to be a problem for the duration of a song or recording session. One could also use an external tmecode box to keep them in sync.
Great concept and video! I would like to see the additional video you mentioned about handling the files going into the daw. I would probably go for one of the portable recorders that also have built in mics-would those work for this as well?
Thanks for the feedback; working on that. You could use a recorder like that as front of kit kid you wanted to. Or experiment with placement. I’d try to get at least 3 mics on the kit if possible, but depending on the sound you’re going for, it could work.
How is the syncing of two units working for you? They are not in true sample accurate sync because there is no wordclock or digital sync between the units, so this would not be the recommended practice of splitting recorded channels between two devices. LTC only syncs the starting point of the recordings, and the units could drift out of time, esp. in long takes.
It's a bit of a hack, actually! I could make a quick video about this, but I used a TRS to dual TS splitter (tip goes to ring on one end and tip to tip on the other) to connect the time code output to the time code input of the other recorder. Since a single jack is both the output and the input, you have to flip the tip and the ring when going from one recorder to another. I then configured one to send timecode and the other to receive. So the two units are sample-locked and the child unit gets the time code from the parent. I still run into some issues where the child is offset by some fixed value from the parent sometimes, probably because they weren't designed to be used this way, but that's easy to fix in post. I actually just ordered a set of TC generators (because of your comment 😂) so I can set them up properly. But for the past 2+ years, this hack has been serving me well!
@@simonpeck I understand what you are saying, I have used Zoom F6 with other devices in the same way. But LTC is still not sample locking like Word Clock or digital sync is, and can drift even when synced with external timecode device. There is no way to sample lock these devices, as they do not have the connections needed. But if it works for you, and you are happy with the end result, it is all that matters!
@@semilumi Oh, interesting. Thanks for the info! Would you be concerned about drift over the duration of a 1 - 3 hour recording session? I ordered the Diety TC-1 set to try out timecode sync using these devices. Should I expect better results that way or would you still be worried about drift?
@@simonpeck The key is the length of the recorded take, not the length of the session per se. You can measure the drift between units by recording exact duplicates of a click track for example. Let that run for the max amount you think your takes would need to be. Export the files to your DAW and line the first clicks sample accurately. Look up the last click and zoom fully in, and measure the separation in samples. If there is none even in an 3h take, the F6 units have outstanding Internal clocks, and it is actually no thanks to the timecode connections made, be it F6's or Deity's. The external timecode devices will not make a difference in drift performance. Look up videos or articles explaining difference between wordclock and timecode to get a better understanding of the concepts.
There's one "issue" with 32 bit floating point. That is, it isn't as precise as 32 bit fixed point at representing values. All the fixed integer values exist in fixed point, but comparing floating point side-by-side and you'll notice some integer values are missing. They're unable to be expressed using floating point. However, the benefit from 32 bit float comes from the insane dynamic range and the fact the lack of certain integer values is totally imperceptible to us. But yes, I get your point about why they didn't bother using 32bit integer instead. Just a guess, but I think it might have something to do with the fact modern general purpose CPUs are optimised to process floating point values more efficiently, or perhaps it's something else. Thanks for the vido. 😊👍
Man this makes total sense! Why spend time getting the mix going if you can spend that time getting the performance going and mix it from the widest possible palette. Kick ass video!
You are really on to something here. I think that for a Non-Professional, your investigation and understanding of 32 Bit Recording is outstanding - and I applaud you. I have been involved with recording (Classical recordings) since 1964 when I was 10 years old. I have been full time employed as a Professional Engineer for 46 years. Full time for me = 80 to 110 hours per week. I have some thoughts regarding what you are discussing. A major factor in recording is the Quality of the Convertors. With the Zoom it is like you are photographing on to the highest grade Film - but with a substandard Lens. This is exacerbated by substandard mic preamps (the Zoom). In my humble opinion it would be better to record PCM 24 Bit with great Mics, Preamps and Phenomenal Convertors than the other way around. The chain is only as strong as..... I personally use an iZ Radar 24 Recorder (for 22 years now) which has stunning convertors. Plus top grade Solid State and Tube Preamps - and very fine new and vintage mics. I have tested simultaneous mixes with Golden Ear Major Label Artists where we went from Radar through an Analog Console back into a pair of Radar Tracks and at the same time into the highly rated Antelope Orion 32 (32 Bit into Digital Performer). The Orion Stereo mix sounded great - but when we listened to the Radar playback - we were truly blown away. The only difference was the Convertors. Another thought: When you run some of your drum tracks through outboard gear and back into Reaper you have now lost what you were getting from your original F6 recordings. I am sure that you have considered this. Anyway - you are a Great Thinker (I'm seeing Rodin here - LOL) and I congratulate you.
Thank you so much for the kind words, Rafael! I agree with you about the sound of the preamp / converter stage. The inputs on the F6 stage can be set to line level. I've used external preamps with it before, but I'm not entirely sure that it bypasses the internal preamps when doing this. I'm not sure that it can, given the novel implementation of dual AD converters on each channel. I do believe it makes sense to prioritize microphone choice in recording, so I like this approach because it allows a person to shift their recording budget from recording gear into microphones. Yes, I'm aware that running the tracks through outboard gear after the fact loses something, but there are tradeoffs everywhere in recording and I wanted to offer an approach I haven't seen done elsewhere. 🙏
I'm not a fan of the flavor of the sound of Zoom converters, but they're not horrible. That said, I am impressed at the size of your mobile 12-track rig! Great setup.
Cool video - but what if you don't want to normalize in post? Usually I try to get a solid recording with a strong recorded waveform. When that is achieved I do not need to normalize. When normalizing you also increase any noise in the recording.
Normalizing in this scenario doesn’t increase the noise floor, but for a quiet recording it would bring it closer to the target level. A similar thing happens when you crank up the gain. When recording this way, getting a strong output from the microphones will provide a strong signal, which is not a problem when recording drums. Also, these recorders from Zoom and Sound Devices have extremely low noise. The target audience is film production.
Hmmm. aren't you by defintion limited by the SPL of the mic? As usual ? How could you ever get louder then the limit of the mic? It's simple physics, no matter the bit depths. I wish I had a mic that registers 1500 dB lol
Yes, absolutely. As I point out in the video, with the F6, you’re limited by the analog input circuitry. But you’re no longer limited by the recording medium, and recorders with more dynamic input circuitry could be built in the future.
@@simonpeck I got that part, but how much range do you need above a set limit (mic)? I definitely agree on recording a higher resolution and headroom needed for the endgoal, always. But if if the air itself, without the mic, only can go to 190db? Hmmm this reminds me of using an amp with 10x the power you actually need for the speakers. Those speakers are singing like never before. There is some merit in what you say, intuitively. Fascinating.
I’m not sure I understand your point. Clip gain has been a thing for a very long time. 32 bit float, as I understand it, allows you to clip internally in a DAW without actually clipping the output.
Probably wrong: with 32 bits you are able to represent 2^32 distinct values no matter if integer or float representation is used. Float allows wider ranges, but you lose resolution.
Zoom released a 32 float audio interface, use that instead. Attach 1 or 2 mics of your choice to it and record it with whatever device, laptop for example
@@simonpeck if you attach a 5v powerbank to each of them then they have enough power to run without needing to attach them to a power outlet in the wall
Modern way of drumrecording is using the Yamaha EAD 10. If you want a quick work flow ...that is used by many modern drummers. Don Sanderson records his drums 2023 using the EAD10 . www.youtube.com/@DonSandersonDrums
I was really enjoying this video then I realised you’re a small channel, really punching above your weight! Keep it up! I’d love to know more about your whole setup, I just bought my first proper mic (sm57) to do some recording for a project after only using DI bass and guitar and I’m completely hooked. I’ve done some bits at college with their studio but Im looking to get a decent setup going at home.
I actually tried something similar with my Apollo (24bit) and didn’t add any ‘gain’ to it (i think it defaults to 10db of gain?). Made for an interesting experiment. After recording i ran it through actual preamps for fun. Digital and analog. So you can hear the ‘sound’ of the preamp, but something irked me a bit with that test. I think this recorder idea would make it even better. I also have been seeking something to do more remotely and i may just use this if it doesn’t require setting levels.
Yeah, it's really freeing to be able to walk in and hit record. One thing I didn't mention in the video is that the recorder can do 24-bit and 32-bit float at the same time. So you can set your levels the way you normally would for the 24-bit recording and decide which tracks you want to use in post. Also the Sound Devices MixPre 10 II can act as a 32-bit float audio interface if you wanted to record into a DAW.
Subscribed! Great stuff, Simon. I‘m recording my drums with a Presonus Studiolive Mixer as an Interface into Studio One. I have to check out if recording 32bit float is possible within this setup.
"I exceeded the input capacity of my audio interface so I used a pad on the xlr cable." "Levels? Wdm levels I record in 32 bit-float. I don't use "gain" or things like that" The weak link in this setup is definetly the preamps, but it's less about the gear and more about how you use it!
Very interesting! I started learning drums with the same idea. After 10 years I am still not a really good drummer. The more you learn the more you realize that drum playing is almost an infinite task. as a sideway I learned how to program drums. With this I‘m quiet ok. When I record my live drums I mice them with 4 to 10 mics depending on the sound I want to achieve. I have a formula how I process the drums that works for me. Just for the sake of learning: I would be interested how you record the drums are and how you process them.
Very interesting. I use Cubase12 which I can record at 96Khz./ 32 bit floating point. I do this when I record projects that need to have beautiful detail in tone, reverbs and FX in general. Generally not loud rock tracks, as the detail isn't useful. For them I use 48Khz / 24 bit. I leave this comment only as a reference. I hasten to add: The song is the thing. Good songs well played tend to make good recordings. All the Best.
Whatever interface you are using it can pretty much be guaranteed that the A/D converters are not actually 32 bit float. The converters are probably 24 bit but the data can be stored as 32 bit values. Whatever format the values are stored doesn't actually translate into a better resolution than the number is being represented by the bit-width anyway. Hence why the industry still uses bit width to define the resolution of recording and not say use the internal storage representation. For example, the numbers could easily be stored in 64 bit float but that wouldn't increase the resolution the audio was recorded in. Once the data is in the computer it makes sense to have the highest practicable floating point representation hence internally DAWs (like Reaper) use 64 bit floating point. Note, even at 16 bit resolution you probably have more than enough dynamic range to record things like drums even allowing for decent headroom to avoid clipping, -18 dBFS is recommended usually, which would give you 80dB of dynamic range!
Zoom and Sound Devices implemented it by using dual 24-bit converters on each preamp. The problem this solves is that not only do you not need to set levels, you can't set levels. It isn't about having enough dynamic range. It's about eliminating a step from the recording process. You can set the level in mixing, after the fact.
@@simonpeck I see. It's not obvious from the specs of the device but if it has dual 24 bit A/D's it might mean it has 48 bits of resolution. This would mean you actually have a massive dynamic range. The would then just set the mic preamps at the lowest levels to gives some level which is above the noise floor.
It uses one set to high gain and one to low gain and has to make an intelligent decision about which to use and how to combine them where they overlap.
I’m interested to know how your workflow works with this? For example if you wanted to play with a click or scratch track how do you do this? I’ve been interested in a concept like this since I got hip to the sound devices line of products as I’m in the same situation with my home studio being somewhere that I can’t track drums in. Thanks for this stoked to see more!
So how do you sync the drum takes to the music? What about multiple takes of the same song? Audient do an interface that does auto level for drums. You just play a bit and it figures it out.
Playback and/or a click track can be sent into one or more of the channels in one of the recorders. Or it can be done manually. Multiple takes have to be done serially with this method, so that is a potential drawback. They can then be handled in the DAW.
@@simonpeck yeah ok. Seems like a lot of extra work to me. If I was recording drums remotely I would be getting a 16ch EVO and a laptop and using my phone as a remote. I do love finding great non studio drum spaces to record though. Some of the most legendary drum sounds have it been recorded in studios.
The Sound Devices MixPre 10 II can also be used as a 32-bit float audio interface if you wanted to record into a DAW. And yes, this is a great method to record in unusual spaces.
REAPER records at 64 bit float by default. I know from skool that 64 is more than 32. This is terrifying. Is my machine gonna blow up? Or maybe my tracks will mix themselves? Help
What do you mean run the tracks through "preamps later"? You can run 16bit audio through outboard gear later, you dont need 32bit float for that. Also the normalizing can be done the same. The noise floor of mics and preamps themselves limit the dynamic range, and all this 32bit float saves you is the trouble of setting a gain, right? Maybe explain how it's different than a decent 24bit recording setup, besides saving 1 minute setting gain. Have you recorded traditionally before?
It saves a lot more than a minute when you’re using 12 mics, especially when you’re the musician and the engineer. It removes an opportunity for human error. Yes, I record vocals and other instruments traditionally in my home studio. I didn’t mean to imply you can’t run any audio through outboard gear later. The difference is that if you’re in a situation where playing drums is prohibitively loud to play in your home studio, you can record them remotely in a simple fashion (without even using a computer) and handle the rest of the workflow in the home studio.
@@simonpeck thanks for your reply, I guess I'm still learning about this so it helps. I do get it from a workflow stand-point, was really more curious about if there's technically an advantage in terms of dynamic range/resolution. No need to explain further, I've since read up more about it. Rock on...
🤔 32bit float, eh? Well, I've been in that wagon since Cubase VST32 1998 or 99. ✌️😎 The original 32bit float DAW since last century. Never had to worry about my mix clipping and also provided none destructive editing processing. Everything you do in Cubase or Nuendo is on a 32-bit float audio engine. Now everyone is trying to catch up on this.
Most DAWs use 32 bit float internally. But that is different than capturing the audio this way. If you record through a 24-bit audio interface, you’re limited by that no matter what format you’re working with in the DAW.
Hate to say it but.. In this context, 32 bit, float or not, is just a number used for marketing. Due to the noise floor limitations of the preamp (as you touched on) you’re not gaining any more usable dynamic range than a regular 24 bit ADC. And trust me, 24 bit is already a TON of dynamic range. 32 bit is more useful for post processing “in the box”, or in digital mixers. This allows extra headroom to prevent clipping from oopsies in the mixing process, AFTER the signal is digitized at 24 bit. I would suspect you could use the same workflow as your zoom setup, with a 24 bit system, and achieve the same results. Which is fine if it works for you! But 32 bit isn’t achieving anything fancy. For a portable setup, I would usually use a portable digital mixer as my interface (Midas MR18, or a soundcraft UI24) These function as 16+ch interfaces to connect to a laptop, and allows easy remote preamp control via an app. This way you’ll be operating in the sweet spot or your preamp, and getting maximum actual dynamic range.
The marketing is confusing, because it’s the dual preamps and converters that differentiate the recorder more than the data format. Sure, 24-bit has enough dynamic range, but this method eliminates gain controls altogether.
@@simonpeck Oh, so 32 bit is just used as a method of containing the dynamic range of dual-preamps. That makes sense- then you can get 24 bit quality after digital gain later. That’s actually kinda neat. 32 bit is definitely not the hero then, although is albeit more space efficient than recording two separate 24 bit files, one for each pre. Regardless, I enjoyed watching through the entire video, and understanding how floating point works. Thanks!
Mmm….kinda. You don’t have to worry about levels WITHIN REASON. Not just willy-nilly. Prince set an incredibly high bar for hit songs, and the drums were all lowly 8-bit samples from his Linn LM-1. Do really good drums mic’d with great mics 👉potentially👈 sound really interesting? Sure, maybe. In some instances. It shouldn’t be overlooked how much your smart decisions with what you know about drums and mics play a role. Do this with lower bit recording and it will still sound 95% just as good. At least. I can do a different smart thing - I can take cheap drums and put brand new inexpensive Ambassador drum heads on them, and then use nothing but cheap Shure SM-58 mics and I get killer recordings even just at 44.1khz 16-bit, (CD Quality) because I know how to best tune the drums and where best to place the mics. Note: Put distorted electric guitars over ANY of this and it’s all OBLITERATED. That’s precisely why we love distorted electric guitars!! 🤘😆🤘 Using 32-bit float just means we have more headroom. So we can worry LESS about the noise floor creeping into our sound, and we can worry LESS about hitting something too loud and causing digital overload breakup. If the sound you create is crappy sounding, all it’s going to do is give you a super clear recording of a crappy sound. Clearly a lot of research and learning was done here in order to come up with this very clever and very effective recording process. I’m just arguing that one can use this type of a setup whilst recording at a lower bit rate, and the results will be nearly indistinguishable for most listeners. I personally do not find 32-bit float to make much of a difference in the recording process. But in the MIX process it’s a game changer. The DAW called SONOR(now called Cakewalk and given away FREE) was using 32-bit floating point processing for its internal processing and mixing YEARS before other DAWs were, and this was what gave it such a sonic advantage. Is 32-bit float better for everything? Mmm, not necessarily. But it’s never worse. Is 32-bit float just better? YES.
Of course you need to set the levels and sound, how else are you going to make sure the mics are in the correct place? Get it right at the source rules apply in all forms of recording. 32bit float makes no difference to the process.
Recording in 32 bit float changes the process because (not only do you not need to set levels) you can't set levels. The entire dynamic range within the limitations of the analog input circuitry is captured. In terms of mic placement, it's still possible to monitor the output. Nothing about this process goes against the philosophy of getting it right at the source.
How is this honestly different than using an interface, preamps, etc in a “traditional” setup, and recording at 32float? You’re just replacing the interface and preamps with your zoom. Maybe I’m missing something.
The main differences are: 1) It saves time -- There's no need to set levels. 2) It reduces the risk of human error -- There's no need to set levels. 3) Drums can be recorded in a remote location without even bringing a computer. The home studio can remain at home. 4) More of the recording budget can be spent on drums and microphones where the impact will be higher than on gear further down the signal chain.
I see you're really trying to understand this stuff, so here's some extra info for you.
There's a limit to dynamic range at several places in the recording chain. Starting with the mic, the maximum SPL level is a physical problem and exceeding will result in physical damage. The microphone will output a voltage level in a certain range. This is then (usually) fed into a gain amplifier. This amplifier is used to keep your maximum SPL in the voltage range of the next step, which is the Analog to Digital converter (ADC). The gain amplifier has an electrical limitation on it's maximum voltage as well, and passing it will cause clipping or saturation depending on it's construction.
Next the ADC takes the voltage provided by the gain amplifier and converts it into a digital representation. This is designed to convert a specific voltage range, and only that voltage range. This is an electrical engineering thing, and there's basically no way to measure any possible voltage. This range is designed to be good for a unity gain amp stage, and designed with the electronic noise present in the previous stages and it's own circuitry. Typically, on the best audio IDs, this will give a signal to noise ratio of 120db. Exceeding this range will turn on clipping lights or, in the case of Zoom, max SPL exceeded warnings. If you're exceeding the ADC max voltage, you're expected to turn down the gain amplifier or otherwise attenuate the signal. The gain is also used to put very quiet sources more fully into that range, though less so.
As the (pulse code modulated) ADC is designed to measure over a specific value range, it is able to convert the ratio of voltage/(voltage max) into a corresponding digital range. Here's an important point. ADCs are signed integer based, at a fixed internal bit depth, i.e. 32 bit, due to how we measure voltages in electronics. This is basically a ladder of resistors that the voltage climbs, each step indicating a certain minimum voltage level. They will produce an integer over the range of -(2^bit_depth) - 1 to 2^bit_depth. If you want less bits, or a different format, than the internal representation the result is dithered and converted into the output type requested. The also tend to sample at several times greater frequency than requested and resample the output for reduced aliasing. This is the first point where floating point can appear but it's not the fundamental measurement being taken, and this is important.
This means that floating point is not fixing anything about gain staging up front, and it's not better than the original integer measurement. It may even be inferior for storage from a strict perspective as floating point is actually quite bad at representing some numbers. You can read all about the weirdness that is IEEE 754 if you want, but it's not that relevant here. After the ADC is done it transfers the PCM waveform to a computer, file, or what have you. In the case of handheld recorders, it's going straight to file. In your case a 32bit float WAV file.
You take your file, you put it in the DAW and then you process. Here's where a bit of history comes in. Computers used to be pretty bad at floating point. They used to be pretty bad at 24bit integer math too. Even switching a Pentium between integer ad floating point registers (the CPUs internal working space) was a bad performance hit. Early computer sound processing did all it could to avoid that. Most things didn't default to floating point processing, and didn't even default to 24bit processing. 16bit integer was pretty standard on computers, and digital hardware was often using 8bit and even 4bit far more than you'd expect. This meant that even mixing down tracks was so ridiculously headroom starved that anything other than meticulous gain staging would result in huge noise floors.
Thankfully times have changed, processors have changed, audio engineering has moved on. 24bit integer is the default capture format because it's compact, fast to process, and covers all the realistic SPLs we'll ever get past an ADC. Additionally DAWs *all* universally process audio as 32bit floating point after they load a file regardless of what the file format is. You can probably even find a setting to process your audio as 64bit double precision floating point if you'd like. But even loading an 8bit wave file, the DAW will convert everything internally to 32bit floating point. This means loading a 24bit integer wav file is equivalent to loading a 32bit floating point one at the cost of space, and possibly compatibility.
The history of computer audio means that some things are more compatible and universal than others. Floating point processing, having come in relatively late, being not exactly native to PCM, and taking 4 bytes to store instead of 3, means that some waveform editors (The standalone kind like audacity) may not support 32bit floating point WAV files. Most will, as will most DAWs, but the testing on them is less thorough as they're used less overall. Standalone versions of plugins may also struggle here, as this is just not the expected format for WAV files. The same is true of VST plugins supporting high sample rates or double precision floats. Good ones will support, most made with good libraries will work, but some may have odd problems. Why take the risk?
Floating point processing is and was a major coup for computer audio processing, but it's not needed or even advantageous in recording or storage. Set your mics gains' so they don't unintentionally clip, and you're fine.
A side note as a programmer: In computers, integers aren't the same as fixed point or floating point numbers. Integer numbers have perfect precision for the set of integer numbers (the kind without fractional components). Fixed and floating point numbers both can represent the set of real numbers with imperfect precision.
Thanks so much.
Jesse, thank you so much for taking the time to go into such detail. I’ve read through this a couple of times now and thinking about the implications. Thanks for clearing up my conflation of integers and fixed point.
Have I fallen for a bunch of marketing crap? It sounds like there are tradeoffs on the recording side (particularly the further a value is from 0 whether negative or positive?), but is the problem mainly on the processing side, after bringing the file into the DAW?
At the core, what is the impact on the sound, if it’s even possible to put that into words?
@@simonpeck So I looked more into what Zoom specifically is doing as I wasn't aware they were going so hard on 32bit float. They're doing some nice engineering things, but also being a smidge overzealous with the marketing.
Zoom is making their things with _two_ ADCs. They set one to high gain and the other to low gain. They got rid of the adjustable gain amp, instead tuning the gain of the ADCs for the max SPL of the microphone. Gain controls it gives you are digital amplification post sampling. The ADCs are still regular ADCs, but the signal is processed by a microcontroller to produce the final measurement. In float recording mode, they use the signal from both to determine if something is above or below 1.0. In integer recording mode they _still_ use both ADCs but they use the best ADC measurement for the signal at the time to keep it from clipping. They don't say exactly how they're combined, but this is my guess. They could also be doing the same thing as some oscilloscopes and splitting the input range across both of them. That would give you incredible dynamic range at the expense of lower normal signals. That's not really a problem with SPLs of 140 or less though, which still fit into the 24bit space. The design reduces the noise floor so this all works out well.
This is pretty slick, and a good example of a well tuned system from an industrial design perspective. It's not something you couldn't do yourself by recording a signal at two gains, and it still has compatibility problems, but it's neat and well packaged. The magic is really in the dual ADCs though, not so much in the 32bit float.
Things to still note are that even in float mode you _can_ still clip the ADCs. It's just much harder. You also get the same full dynamic range in 24bit mode.
If you're recording to files, the 32bit float WAVs are still a possible compatibility problem, but you''d only need to normalize and convert them with something to use them anywhere.
If you're using it as an audio interface, then you may have issues recording directly in 32bit float. Zoom has a compatibility table, and I'll point out that it's probably also only valid with ASIO drivers in Windows. WASAPI does have 32bit float format support, but it's pretty unused and thus not tested well. You might end up with not being able to record it, or recording it but any peaks past 1.0 are clipped, or other strange bugs. I would test anything before using it for anything important.
As you get the same not-have-to-set-or-worry-about-gain in 24bit, anything that doesn't behave well has a good fallback.
jesse, you been smoking some AI?? just kidding. great info!
The noise of your analog input limits the dynamic range to about 110db/20 bit at best. 24 bit recording with levels set in the middle, with plenty of headroom should be good enough then. The 24 bit master can then be used to create a 16 bit file for listening... still with a 90db dynamic range (which is huge for popular music).
There's no question that 24-bit is enough. The purpose of this techniques is more about how recording in 32-bit float can change and simplify the workflow, especially when it comes to recording with a dozen microphones. It isn't an argument that the dynamic range or sound quality of the final product is better.
@@simonpeck Hey Simon, thanks for replying. Great answer!
Interesting video! Was hearing a bit about 32-but float about 10 years ago and people seemingly stopped talking about it. Happy to see such a good application for it with a straightforward explanation.
Talk about an underrated channel with amazing quality, explanation and attention to detail. Dude keep it up, its amazing
You're too kind!
I was just wondering, how can this channel only have 400 subscribers?? That will be over soon, because this is exactly what is good content about 32 float
Really enjoyed this video. Very professional. Looked down at stats. What 40 views??? Where are the people!??
Haha! I think you jumped on right after I posted it. Thanks!
I appreciate the segmentation, straightforwardness, and overall flow of the video. Your thoughts and opinions feel very grounded to reality and you support those with solid information that you've acquired through experience and ultimately by just being familiar with the equipment in question. Keep up the great work, you've most certainly earned yourself a subscriber.
Thank you so much! 🙂
5:34 the converters aren't actually 32-bit float, they are 24- or 32-bit fixed. The floating point format is something completely digital and can't be recreated in the physical world. They instead use two, or three, converters at different gain levels and the 32-bit float files are then calculated from the combined signal of the converters. This is why you can always recreate the waveform even if one of the converters clip, you just replace the clipped part with the signal from a lower gain converter (although it's a bit more involved than that).
Yes, but that is an implementation detail of the Zoom F6 (and other 32-bit float recorders currently on the market). For practical purposes, it doesn’t matter. As a recording engineer, it doesn’t affect the approach to recording.
Thanks for that hint, that explains to me why the whole thing is working in the first place.
From that information it is clear that creating gear that covers a larger range is not impossible but will be exponentially more expensive because for every increase in dynamics you need to add:
A) another gain stage + ADC
B) more processing power
C) more internal bandwidth
D) possibly another storage device because at some point your SD card will bottleneck
Also, there is not only the upper limit but a lower limit as well, which is mostly the preamp in the Microphone, as well as thermal noise in cables, resistors and so forth, which at room temperature and a bandwitdth of 20kHz is roughly -130dB. There's also something like bit flicker and clock noise in the converter, so the real dynamic range of any device is much less than what's theoretically possible with 32bit float, even if you split the signal across multiple amplifiers. Also, unless you are in a well treated acoustic room, the normal room noise kicks in much earlier than the thermal noise. At >60dB of amplification, it usually becomes quite audible. Which is to say that you will never get rid of the task of getting a decent signal to noise ratio in the first place. But then again, drums are a very loud instrument, so you get a good signal to noise ratio with almost any recording setup (unless you play very dynamical with extremely quiet drumming mixed with normal/loud drumming). With drums, any modern AD-converter should give you a good signal to work with. In my experience, the most important factors are: Mic placement, mic choice, acoustic sound of the drums, acoustic sound of the room... and the rest is absolutely minor compared to these four things.
@@simonpeck N J has a valid point. On audio interfaces that use only one converter per input channel the converter is always the bottleneck for the signal. It can be 24 bit or 32 bit but it is always fixed-point, so you lose any advantage that floating-point offers already at that stage. Zoom tries to overcome this bottleneck by using two converters in the F6, each with its specific dynamic range. Your example with vocal and drums is basically what Zoom describes in their manual.
So the first answer to the question raised in your video title: Not everybody records drums like this because not everybody has a multiple-converter interface.
Your thinking of recording "dry" and do anything else later in the box was already envisioned by some producers in the 1980's. I saw a documentary about Trevor Horn's production techniques. They tried to do "in the box" whatever was possible to do in the box back then.
That leads to the other answer to the question: Plenty people working in music studios still behave like in the analogue days.
@klauba Yes, you’re right about the bottleneck. You must have a 32FP recorder or interface to do this. My question was how come “nobody” records drums like this, not “everybody.” I was sure others must, but I hadn’t seen any other evidence of that online. I was hoping those who do might be drawn into the conversation, and a couple of people have. 🙂
I have been recording in many cases where all I had was an interface and 8 channels ADAT with the equipment in the same room as the instruments being played, Thus not being able to make and compression or EQ decisions, They would be left out of the signal chain altogether. Sensibly I would set gain levels to be safely below clipping and proceed to do up to 1- 1/2 live takes, All the decisions had to be made later. I found 24 bit was perfectly fine to get good clean captures , even from substantially boosting quieter parts. What a great time to be making music in! and great vid!
Thanks for making this video, there have been many times where I have had the opportunity to record in 32bit float but chose not to due to my gear “not being professional” by some opinions.
You have inspired me to try it out and work with the post record workflow you described.
I’d love to see your remote setup in a future video
Yes! Try it out. What type of audio are you recording? There does seem to be some trepidation around the topic, but don't let that stop you. It's freeing to know you don't have to worry about levels. And in the case of a drum kit, it's easy to make a mistake when setting the gain correctly for 12 microphones.
@@simonpeck I mostly work on recording my personal music ideas. Also, I am starting to get a little more serious as a sound designer. Finding sounds with a field recorder and not “needing” to adjust the gain constantly will be a timesaver.
floating vs fixed as a choice for digital audio is somewhat analogous to the use of dolby in analogue audio, in that it is advantageous to s/n ratio when used properly, but goes horribly wrong when it's not; bit-rounding that can happen to digital signals if mishandled will give unpredictable results because vital parts of the number are lost. this doesn't happen with fixed point, as you say.
I think this reasoning is part of why 32-bit float was chosen as the format rather than 24-bit float.
Amazing, very informative video thank you for the post. I've been recording in 16 to 24-bit wav audio over the last 25 years. I have never had enough money to buy a 32-bit float system. Recently using a rode podcaster i used the the 32-bit float option. Like you mentioned, I was able to change all of the levels after in post from tiny little laughter's too loud comments during the podcast, game-changer for sure.
Oh nice, I didn’t realize the Podcaster supports 32-bit float.
making electronic music for the past 15 years I never thought about using 32 bit recording/files (mainly because any genre of electronic music is rooted in old, far less capable samplers than the machines we have today), however, this was such an amazing and interesting conversation on the whole topic!
way underrated channel. think this is gunna be the video that pops for ya. keep up the great info an quality, its highly appreciated(sub 200)
Thank you!
I’d love to see how you set up everything to record the drums since I’m looking into how to record drums as well
That's great to hear! I was hoping someone would want to see that. Stay tuned...
Thank you all for watching this video and your incredible, detailed, well-articulated comments. I'm blown away by the response!
CORRECTION: In the video, I mistakenly said 32-bit float allows for 4.6*10^38 "states." What I should have said is that it can represent numbers within that range. There are still 2^32 states. Because some bits are used for an exponent, there are fewer available for the mantissa, meaning 0s are added, resulting in a larger distance between values at the extreme ends of the positive and negative values.
Nice subject. Recording is more than mp3, 16 bit, 24 bit or 32 bit. Musicianship, song, arrangement, production, instruments, acoustics, microphones, mic placement, pre amps, mixing, mastering, exporting. Audio/video. Light. Regarding recording technique: Noise of electronics, clipping, noise of drums, tuning, noise from outside, noise or degradation of signal in software or signal chain. noises in room acoustics, etc. etc. Phase correction can improve significant a recording. Noise really is irritating and can destroy a recording however sometimes a noise gate can help and adding some effects like subtractive eq sometimes also helps. There also are software type noise cleaners. Multi track is often more intensive than just mono or 2 mic stereo. Adding a lot of data types together like several audio tracks and several video tracks can take a lot of time for the alignment. Recording on pc/laptop does not require the sd card and sometimes it takes a lot of time for datatransfer and ...have a rough mix available at the time of recording (pc/laptop) may help to improve the artistic or technical performance during the recording session. On the other side: many people use recordings of iphone or smartphone and ...after mastering with iZotope that can sound OK for the purpose: showing you can drum or had a nice gig, a nice band or have made a nice song or just had fun.
@@HenkJanDrums Yes, this is all true. In my own recordings, I rarely, if ever use all 12 tracks. I'm always swapping mics, trying out different positions and sometimes mic the same drum two different ways. Later, I listen to how it sounds in the mix to decide what to use. Phase is often the reason to leave something out. But sometimes it just sounds better without certain mics in the mix. The target style for a particular song is another major factor.
That was awesome please show us your whole Process of finishing the drums in the daw
i never knew such a concept existed, though intuitively i always sensed that levels should've been licked by now. great vid, man.
It’s a novel solution and a pretty good one, but I don’t think I’d call it licked yet! That could be just around the corner…
Recording (in terms of converting from analog to digital) is always done in fixed integer (16/24/32 bit). You MAY choose your recording-file format to be 32 bit float, but the dynamic range in that format is also only 24 bit, because of the nature of the IEEE-754 floating point format. You will even get more dynamic if you record in 32bit fixed integer, because then you really have 32 bits of dynamic range, BUT, this is really over the top. I don't think, that someone can hear the difference of 32 bits vs 24 bits dynamic... You will even hardly find AD-DA converters doing real 32 bit conversions seriously, and very often the internal signal chain of plugins is also only 32 bit float (=24 bit dynamics), so, recording anything higher than 24 bits is practically more or less useless, most of the time.
In 32 bit floating point, you have an effective dynamic range of 24 bits. Other than in integer formats, you can not clip the value range of float samples that easy, which maybe is, what you meant, BUT, always keep in mind, that even if you can produce values outside the range of -1.0 to 1.0 during mixing (and also when rendering to a file), there comes the time, where the data wants to be translated from digital to analog, again. And this is the moment of truth. Everything outside -1.0 / 1.0 will be clipped. Maybe some players have intelligent built-in limiters, maybe.
So: If you like the "sound" of 32bit float much more, then you most likely like the "sound" of your limiters.
To really have MORE dynamic, you should record in 32 bit fixed integer, and put your DAW into 64 bit floating point mode (if that is possible). THEN (and if all your plugins support 64bit audio processing) you really will have more dynamics during mixing. But whether your monitor / your ears will hear the difference, I don't know, but I would bet, that (especially in home-recording scenarious) you won't hear ANY difference.
You captured what I was going to say in more detail and clarity than I could have. Thanks for pointing out what I think is one of the key sets of facts.
Not in this case! The Zoom and sound devices recorders use dual 24-bit converters on each channel. One handles high gain and the other low to record in 32-bit float. When choosing this format, the limiters are bypassed.
@@simonpeck
I have to admit, that I was not aware of that technology... And it indeed sounds cool, in terms of "setting proper recording levels, that do not clip".
While many of the things I said are still correct, I was really wrong about the dynamic regarding capturing the audio. I am sorry for that.
Obviously, the sampling process produces 2 different values per sample. One high gain and one low gain, then the high gain one clips earlier (depending on the gain differences), and still a valid resulting sample value can be reconstructed by using the low gain value, to reconstruct the clipped sample.
And yes, in this case you can have more dynamic headroom, depending on the overlapping range of both gained converters, and I was not aware of that. But on the other hand, you really need to work on that recordings (as you showed in your video)... and for ONE thing it is great: you do not have to care too much, to get the right recording gain level.
Which indeed is somehow "cool"...
What about the noise floor?
Exactly what I thought
would love to hear this answer👆
Are we thinking of having a mezzanine?
My thought exactly - not only there is a noise floor on equipment, there is an “acoustic noise floor” too - the quietest studio will always have some kind of background noise.
I didn't hear any noise in the whispered vocal in the video.
Preamps and other segments of the audio chain up through and including the ADC are not necessarily linear. This approach MIGHT work well with awesome preamps, ADCs, mics, room sound management, etc., but even bleed will add pressure to the components in the audio chain and color the sound. And if you set the gain too low, you’ll have analog noise to deal with when you raise the levels of the channels you want louder in the recordings.
In a perfect and perfectly digital world, this is a great way to skip over some of the work you’d normally have to do in setting the gain for recording drums. But, this is not the world we live in.
Well, how does it sound? I’ll try to post a better example in a future video.
This was a very interesting video.
I think some clarification of the pre-amp stage vs the AD conversion is important whenever talking about this advance in tech. The max input level of 131db is the point where the analog circuit before the conversion runs out of headroom. Unless a person is strictly discussing a conversion only unit, that limitation is going to be an issue with any system offering 32BF conversion. Then again, if you're working with converters that do not include pres, you're probably well beyond any discussion we could have here. Eventually, we may actually be able to record audio in a RAW format (like images and video). Exciting stuff.
New sub. Thanks.
I've noticed that the same limitation exists across manufacturers. Can you shed some insight as to why that is? Why couldn't someone make a pre-amp stage that could handle, say... 150dB?
Wow! Outstanding info! A lot to digest. Thank you!
Dynamic range isn't the only consideration. Virtually all analog gear have a noise floor that is audible. Raising the source level relatively high above that noise floor ensures a better signal to noise ratio and a cleaner sound.
Very enlightening - thank you
Interesting, may have to give this a go
Finally. Good stuff. TH-cam has some new dudes killin it. Just great info. Solid
Thank you.
Generally a great video about an often overlooked topic. Here's a few things I'd like to point out tho:
1st) Regarding the maximum SPL, I don't see a reason why there should be a limit on anything that isn't a microphone capsule. All the microphone does is convert the atmospheric pressure waves (SPL) into a voltage. The higher the SPL, the higher the voltage output. This is heavily influenced by the sensitivity of the microphone. So, for a less sensitive ribbon microphone it would take a much louder sound (higher SPL) than for a sensitive condenser microphone to output a voltage that's higher than the mic pres can deal with.
2nd) As some commentators here have pointed out, making use of the entire theoretical dynamic range of 1528 dB is technically impossible. However, neither do those recorders simply store the data in a different format nor does a 32 bit converter chip even exist. What's really going on inside those units is a clever combination of fixed-gain mic pres and regular converters. Think of it like running your microphone signal trough a signal splitter and feeding it through multiple preamps with different gain settings in parallel. The outputs of the preamps then gets digitized through regular converters and are digitally stitched together. You're only using each mic pre at/around its ideal operating point. This way your gain setting will always be correct, which makes 32-bit recorders an excellent choice for any situation where levels are changing rapidly or are completely unpredictable.
3rd) I really like your idea of recording the drums as cleanly as possible and then adding any coloration later. But 32-bit won't really offer you any higher fidelity than a regular 24-bit recorder with properly gained preamps, as the fixed-gain mic pres still have a noise floor that won't be any lower relative to the operating point aka your signal. However, I can totally see how not having to deal with setting levels can make the whole operation a lot easier, especially when doing everything yourself. 👍
Here's a better explanation than I can deliver in text form: th-cam.com/video/NTuJ1fk3PsY/w-d-xo.html
Thanks for your comment! Yes, agreed with most of this. The only thing that I would point out is that most microphones have a higher max output than the max SPL of the recorder’s input circuitry of 131dB. Since you mention ribbon mics, one of the mics that overloads the F6 is an active ribbon mic on the kick head. I put a 10dB pad on it.
Also, I’m not claiming higher fidelity, but rather an easier workflow with a recording that sounds good.
I intentionally left out how this specific recorder works, because I wanted to talk about the principle of recording in the 32-bit float rather than focusing on any specific manufacturer’s implementation. However, since there seems to be a lot of confusion about it, I may make that video.
Great explanation. I use a sound devices mix pre 6 ii and 4 of the slate ml2’s. It is such a great combo to set up my drums in other places and record. Then bring into the big system.
Nice!
Excellent description of 32-bit float and your drums sound great. Keep 'em coming!
👊
That's a really cool way to do things. I definitely want to hear more about it
Did not understand all the tech-babble (yet), but sure sounds like some innovative way to record/mix drums (I'm a drummer myself), ... so kudos to you and all the time you spent to come up with this technique and even made a vid for all of us to learn about it. A+ from me.
very nice setup. thanks for the info! ;)
You bet!
You solved a problem I didn't know that I didn't have..
yes in your case you can save some money. thats a good reason to do this. otherwise there are other downsides involving the double-AD converter design and other stuff that is impossible for amateurs to discuss. you can read about that in bob katz mastering forum or in discussions including Lavry Engineering or daniel weiss, which are some of the best converter designers. btw stagetec has done this for years on a professional and very expensive level, but they didn't sell many consoles.
Nice mix of video production tips and audio production tips. I'm more interested in the audio stuff personally but always nice to know how to improve pictures. Cool channel 👍
Thanks for this. I’ve been going down a rabbit hole learning about 32 but float and trying to decide if it makes sense to upgrade.
For almost 15 years i record mp3 320 kbps up to wav 24 bit with zoom H4, H5, R24, and L20. Live mostly stereo (Zoom H5) and in studio multitrack (Zoom L20) up to more than 16 microphones nowadays. For stereo live recordings I use a modern version of Soundforge Audio Cleaning Lab (24 bit), that has a super fast workflow for making normalised and mastered songs out of a live recording. For home studio Zoom L20 multitrack recordings i Use Studio One professional (64 bit Float) . I record with recorders because they dont have fans like mixers or pc. I record with low gain values and do the processing in DAW afterwards. In my DAW I only work with normalised tracks so ...new recordings are mixed already very well when imported to older projects. With Zoom H4 and now H5 I always have a good live recording.The Zoom gear has almost no noise levels for serious hobby. However some cheap microphones may have noise levels. I can recommend iZotope for mixing (Neutron) and mastering (Ozone) it has automatic (analyse audio) settings that are really great. They can be set to drums but also to snare drum, bass drum, toms, percussion, acoustic music, detailed drums other styles of music, etc. The Zoom L20 has better pre amps than the R24. The whole process from recording to video did 10 years ago give a big loss in the video software. I now use modern Vegas Pro which has very good audio. To show how quality was many years ago and how it is now: see 2 video links of same song, same drumkit, same drummer but Zoom L20 in stead of Zoom R24, better mics, and most of all better video software. All recorded 24 bit ...than and nowadays.
Modern mix th-cam.com/video/xe1couRF5Ms/w-d-xo.html
Old mix 10 years ago th-cam.com/video/L1k9u7dVE9Q/w-d-xo.html v
Modern drums th-cam.com/video/WmFeKEgfyWE/w-d-xo.html
Modern drum time aligned (edited towards 100% exact timing) th-cam.com/video/GdOy_U_tSG0/w-d-xo.html
Same type of soft acoustic drums 10 years ago. th-cam.com/video/uAs0w4DvO5A/w-d-xo.html
Great video! I would love to see the complete setup.
At 5:10, I don't understand the claim about different numbers of representable states for 32-bit integer vs float. With 32-bits it is possible to have at most 2^32 different states, no matter the data format. Integer format uses these states to represent equally spaced numbers, whereas floating point format has numbers that are closely spaced around zero and farther apart the larger they get.
Oops, you’re right. I should have spoken about the range and not the number of states…
It appears then that you are recording the drums "wild" outside your DAW. If so, for me this would make keeping takes organized a hassle as trying to do things like "punching in" a total nightmare. Plus all of those performances would need to be hand synced when laid into the DAW - meaning maintaining the same time placement of what was originally played a crapshoot.
And all to preserve a dynamic range that no drum performance would ever need or take advantage of and avoid the hassle of setting levels - which as someone who has been recording drums in home studios for 30+ years, I find I only have to very infrequently do. Moving from full on rock playing to playing a jazz ballad with brushes would require a level reset - but staying within the range of most rock/pop/country/funk playing really rarely does.
Just seems little (to no) advantage, yet more hassle compared to just having a standard 8 channel interface (or 2 of them) that stay with the drums - with the DAW session on an easily transportable laptop. But an interested thought...
I will say that not being able to punch in is probably the biggest drawback to this method. It isn't impossible to work around that, and because it isn't as clean as when working with a DAW, it does create a little pressure to record a better performance. So I wouldn't say it's all bad.
However, aligning tracks is not an issue. These recorders can be configured to produce a multi-track, "poly" wav file, so all 6 tracks are in an individual file. If you use multiple recorders, there is at times one alignment procedure, but it only takes seconds to complete. In my next video, I'll show you how I work with these files and how I handle handle takes, since you asked.
Since it sounds like you have rich experience in this area, I'd love any insights you have on that recording process. Thanks!
@@simonpeck To clarify, I wasn't speaking of keeping the drum track files lined up (that would be relatively easy with the recorder producing individual files - as they all start at the same time, which when recording a whole set, they would). No I was referring to laying them against the tracks, you are overdubbing to. The drum tracks as a group would be "wild" to the reference tracks - and thus the "feel" of the performance against those tracks as played would not be discernible - basically left to the ears of the person adding them to the DAW. Again maybe not the biggest deal - but personally, I like to know where I played things against the rest of the music, before I start moving things around - either by hand or by timing tools.
Again I can certainly imagine the ways to make this work - I just personally couldn't imagine making those compromises for what seems to me to be so little advantage (if practically any at all)
You asked "why doesn't more people do this?" - and the answer was sort of baked into all of those videos referred to - but then sort of dismissed. Things like not being able to punch, having to constantly manage, transfer and re-align track/files, not being able to easily listen to performances back while recording, having a process that really dictates going for complete takes... all of these extra inconveniences chew up time and energy... for a process that is already left brain/right brain demanding... where every added technical step just sucks away the player's ability to focus on playing and the music... Maybe soon there will be interfaces that will record into the DAW in 32 bit float - then we could have these advantages without suffering the workarounds. But until then, I can't see the answer to your original question being anything other than...."because it's not worth it".
By the way - while I'm sort of busting your chops on this, I've always been enjoying a lot of the other videos on your channel. Good stuff. All the best, David
@@DavidCrigger Oh, I see... I do play against a source with in-ear monitors. I haven't done this, but you've just given me the idea. to try it out: I think that monitor source could be fed into one (or two) of the inputs on the audio recorder. The inputs can be configured to accept line level. That might solve part of the problem, though it isn't as elegant as a traditional approach.
Incidentally, Zoom just released a 32-bit audio interface and Reaper supports 32-bit float recording, so I think we are on the cusp of having the best of both worlds. That was my hope and prediction at the outset of making this video. The Zoom only has 2 inputs, but maybe Universal Audio will add something to their lineup in the near future. I suppose one could get 6 UA232s and create an aggregate device for 12 inputs. 😂
Great video! You explained this concept very well and earned a new sub. Keep it up 😃
Check out a Burl Mothership (either the B16 or the B80). It supports 32 bit float and the preamps are amazing.
In conversion Ad and DA you can still clip the audio. Only internally you are "safe" from clipping.
Yes, agreed: 08:32
Thank you for this explanation. I guess I'm recording drums at 32 bit float from now on! What is the difference in file size versus 24 bit PCM?
The files are 33% larger.
I want a room for drums too. I must use midi drums now.
Yeah, it's nice to be able to cut loose without worrying about noise!
Although there are good reasons to use 32bit float but you can't just ignore everything else. There is still a signal to noise ratio you want to watch and the dynamic range of the analog part of your chain prior to the converters can cause problems. In addition, there can be sweet spots where the actual equipment sounds the best.
BTW seeing as you mention Reaper and recording Drums in this video, can you do another video on how to simplify drum replacement in Reaper, please?
Yes! I'm working on that.
Good vid, good process. I'd suggest, as a guy who used to record live jazz, that you try for a point where you use two mics max to record a kit. W/out deep explanation, one can get full flexibility at mix time with a clear, two mic recording and avoid phase issues, and even the need for processing beyond limiting and verb.
The difficulty of "good drum sound" is reassembling a kit with 8-100 mics. If you can play somewhat evenly balanced, you can avoid a gaggle of mics, and multi-band compression can give you eveything you need at mix with a simple stereo mic pair at MS or XY config...
In short, too many feel the need for too many mics.
Most I've ever used on a kit is 4. Give it a go. The clarity is sweet.
Thanks, Roth! 🙂 I started with 3 mics, used the Glynn Johns method and slowly built it up over time. I regularly try different configurations from 1-18 mics and yes - phase issues can be a headache. I have loosely planned to do a series of videos using a different number of mics each time if that’s of interest.
@@simonpeck Second read of my blathering sounded a little *actually.* I should say "I spent thousands on drum mics only to use 2-4mics."
Sorry if I came across like a net know it all.
Not at all!
All the mention of max decibel levels is great but it really doesn’t matter. What matters is dynamic range. If you’re recording a drum set, you’ll maybe have a what, a 20-30db dynamic range in even the most extreme case? In your configuration, you are still using analog mic pre’s and A/D converters, subject to S/N ratios and clipping. You can still normalize/boost a quiet part of a 24bit (or 16bit) recording. That’s been happening for 40 years. Maybe if you had a a mic that puts out a straight 32bit digital signal (I know Rode makes one) this would make sense, but I can’t think of many scenarios were I don’t want to take a few minutes to get good levels on the front end and do it later.
this
The point is not to celebrate any specific recording format - it’s to make the process faster, easier and less error prone, while focusing on the sound where it matters most: the drums and mics. I want to use my favorite mics for each surface.
Then, good levels can be set in post, believe it or not.
Do the drums sound bad? I probably should have had longer clip of the drums mixed, but I can post that in another video.
For the past 40 years, all of the production hardware has had to be within a cable length of each other. In my setup, the drums are in a completely different physical location.
@@simonpeck you’re title says “how come nobody records drums like this” and I explained why. Yes, levels can be adjusted in post, wether 32 or 24 or 16 bit, that is sometimes called mixing. Not sure what you mean by in a different location. I’ve had various 16plus track mobile recording rig (usually 2-4 rack spaces now a small digital mixer) for going on 20 years now. For me, I’d much rather take a couple minutes run in the zone it runs optimally. Again, analog in means S/N ratio is a factor and clipping is still a possibility at both the mic pre and A/D converter.
@@zagatoalfa Got it. I would love to know more about your mobile recording rig! Yes, clipping is still a problem in my situation, which I talked about at 08:30 - I used a Shure input attenuator on a couple of mics in that situation. What I meant by "different location" is that my drum studio is not in the same location as my home studio. So it's basically a mobile recording setup. What do you currently use for yours today?
@@simonpeck for my portable rig in the way back I’d use a digi002/003, a Behringer 8 channel pre converter combo, and a Sytek MPX 4a for four “money” channels, giving me 16 mic ins. More recently a Behringer XR18 has sufficed for full band location recording. I’ve sold the XR18 to help pay for a Yamaha DM3 that will give me 16 mics into laptop via usb and be generally more useful for my present needs. I’ll note, all of these set ups do require a computer, though there are a number of companies that do multitrack stand alone mixer set ups these days. I did borrow a Zoom F8 in the early days of the pandemic to record rhythm tracks for a friends band in an alley. Battery power was a good option at the time!
Another note about setting levels when tracking… if your levels are super low or all over the place, it’s going to make monitoring what your recording pretty difficult. If you’re recording more than one person, and they are on headphones, and the levels going to your DAW aren’t adequate, it’s going to be more trouble getting headphone mixes sorted than just tracking at a good level. Best way to have a bad day in the studio is to have a bad headphone set up.
End to end process for sure
GREAT video!
Thank you!
If I at some point will once record real drums... I will do it like this. Really great approach
I’ve been drumming for more than 20 years and a software professional for more than 14. I just bought recording equipment for the first time. I’m intrigued by this technique which reminds me of light field photography. I hope my equipment lets me capture in 32-bit float format.
Interesting comparison! I had a Lytro when it first came out. I’m working on a couple of follow up videos (in parallel 😅) to dive in a bit deeper.
Loved the NileBlue reference!
It sounds like your drums were in a room with no reverberation . And the compression is wild!
It’s a small room with a fair amount of acoustic treatment. It has a dropped ceiling with about 12 feet of empty space above it.
Such a great video! 😊
32 bit doesn't sound like a lot. Shouldn't it be at least 44? Or do you like the compression?
Great info here, Simon!
The amount of hiss present in your demonstration further proves that you still need to check your levels.
I probably should have mentioned it in the video - what you’re probably hearing is the combined self-noise from 12 microphones, one of which is a tube and two of which are ribbons. The mic I normalized for voice was about 30” away from my head. The noise floor in the room is 28-30 dB. There is no way to set levels in 32 bit float mode.
Without Wordclock, I’m curious if you notice any issues with sync-ing the two recorders 6-track cards together? Looks like you’re just eyeballing it. I imagine they probably start out pretty close, but long takes will drift apart since Timecode is merely a Timestamp at the head of each file. If you split a test tone like a 1k sine wave source into both F6 recorders and let the take roll for a while, I bet you’d hear (and see) the drift when you line them up in the DAW.
Good question! The drift of the F6 is 1 frame every 48 hours, so if they were drifting in opposite directions that would be 1 frame in 24 hours. I haven’t found it to be a problem for the duration of a song or recording session. One could also use an external tmecode box to keep them in sync.
Great concept and video! I would like to see the additional video you mentioned about handling the files going into the daw. I would probably go for one of the portable recorders that also have built in mics-would those work for this as well?
Thanks for the feedback; working on that. You could use a recorder like that as front of kit kid you wanted to. Or experiment with placement. I’d try to get at least 3 mics on the kit if possible, but depending on the sound you’re going for, it could work.
How is the syncing of two units working for you? They are not in true sample accurate sync because there is no wordclock or digital sync between the units, so this would not be the recommended practice of splitting recorded channels between two devices. LTC only syncs the starting point of the recordings, and the units could drift out of time, esp. in long takes.
It's a bit of a hack, actually! I could make a quick video about this, but I used a TRS to dual TS splitter (tip goes to ring on one end and tip to tip on the other) to connect the time code output to the time code input of the other recorder. Since a single jack is both the output and the input, you have to flip the tip and the ring when going from one recorder to another. I then configured one to send timecode and the other to receive. So the two units are sample-locked and the child unit gets the time code from the parent. I still run into some issues where the child is offset by some fixed value from the parent sometimes, probably because they weren't designed to be used this way, but that's easy to fix in post. I actually just ordered a set of TC generators (because of your comment 😂) so I can set them up properly. But for the past 2+ years, this hack has been serving me well!
@@simonpeck I understand what you are saying, I have used Zoom F6 with other devices in the same way. But LTC is still not sample locking like Word Clock or digital sync is, and can drift even when synced with external timecode device. There is no way to sample lock these devices, as they do not have the connections needed.
But if it works for you, and you are happy with the end result, it is all that matters!
@@semilumi Oh, interesting. Thanks for the info! Would you be concerned about drift over the duration of a 1 - 3 hour recording session? I ordered the Diety TC-1 set to try out timecode sync using these devices. Should I expect better results that way or would you still be worried about drift?
@@simonpeck The key is the length of the recorded take, not the length of the session per se.
You can measure the drift between units by recording exact duplicates of a click track for example. Let that run for the max amount you think your takes would need to be. Export the files to your DAW and line the first clicks sample accurately. Look up the last click and zoom fully in, and measure the separation in samples. If there is none even in an 3h take, the F6 units have outstanding Internal clocks, and it is actually no thanks to the timecode connections made, be it F6's or Deity's. The external timecode devices will not make a difference in drift performance.
Look up videos or articles explaining difference between wordclock and timecode to get a better understanding of the concepts.
@@semilumi Okay, I will do that. Thanks so much for the info!
There's one "issue" with 32 bit floating point. That is, it isn't as precise as 32 bit fixed point at representing values. All the fixed integer values exist in fixed point, but comparing floating point side-by-side and you'll notice some integer values are missing. They're unable to be expressed using floating point. However, the benefit from 32 bit float comes from the insane dynamic range and the fact the lack of certain integer values is totally imperceptible to us. But yes, I get your point about why they didn't bother using 32bit integer instead. Just a guess, but I think it might have something to do with the fact modern general purpose CPUs are optimised to process floating point values more efficiently, or perhaps it's something else. Thanks for the vido. 😊👍
Man this makes total sense! Why spend time getting the mix going if you can spend that time getting the performance going and mix it from the widest possible palette. Kick ass video!
Thanks, Pablo!
You are really on to something here. I think that for a Non-Professional, your investigation and understanding of 32 Bit Recording is outstanding - and I applaud you. I have been involved with recording (Classical recordings) since 1964 when I was 10 years old. I have been full time employed as a Professional Engineer for 46 years. Full time for me = 80 to 110 hours per week. I have some thoughts regarding what you are discussing. A major factor in recording is the Quality of the Convertors. With the Zoom it is like you are photographing on to the highest grade Film - but with a substandard Lens. This is exacerbated by substandard mic preamps (the Zoom). In my humble opinion it would be better to record PCM 24 Bit with great Mics, Preamps and Phenomenal Convertors than the other way around. The chain is only as strong as..... I personally use an iZ Radar 24 Recorder (for 22 years now) which has stunning convertors. Plus top grade Solid State and Tube Preamps - and very fine new and vintage mics. I have tested simultaneous mixes with Golden Ear Major Label Artists where we went from Radar through an Analog Console back into a pair of Radar Tracks and at the same time into the highly rated Antelope Orion 32 (32 Bit into Digital Performer). The Orion Stereo mix sounded great - but when we listened to the Radar playback - we were truly blown away. The only difference was the Convertors. Another thought: When you run some of your drum tracks through outboard gear and back into Reaper you have now lost what you were getting from your original F6 recordings. I am sure that you have considered this. Anyway - you are a Great Thinker (I'm seeing Rodin here - LOL) and I congratulate you.
Thank you so much for the kind words, Rafael! I agree with you about the sound of the preamp / converter stage. The inputs on the F6 stage can be set to line level. I've used external preamps with it before, but I'm not entirely sure that it bypasses the internal preamps when doing this. I'm not sure that it can, given the novel implementation of dual AD converters on each channel. I do believe it makes sense to prioritize microphone choice in recording, so I like this approach because it allows a person to shift their recording budget from recording gear into microphones. Yes, I'm aware that running the tracks through outboard gear after the fact loses something, but there are tradeoffs everywhere in recording and I wanted to offer an approach I haven't seen done elsewhere. 🙏
Thank you
Super video pal, you have a good manner for videos on music gear, good look with the channel
Thanks so much!
I'm not a fan of the flavor of the sound of Zoom converters, but they're not horrible. That said, I am impressed at the size of your mobile 12-track rig! Great setup.
Cool video - but what if you don't want to normalize in post? Usually I try to get a solid recording with a strong recorded waveform. When that is achieved I do not need to normalize. When normalizing you also increase any noise in the recording.
Normalizing in this scenario doesn’t increase the noise floor, but for a quiet recording it would bring it closer to the target level. A similar thing happens when you crank up the gain. When recording this way, getting a strong output from the microphones will provide a strong signal, which is not a problem when recording drums. Also, these recorders from Zoom and Sound Devices have extremely low noise. The target audience is film production.
Hmmm. aren't you by defintion limited by the SPL of the mic? As usual ?
How could you ever get louder then the limit of the mic? It's simple physics, no matter the bit depths.
I wish I had a mic that registers 1500 dB lol
Yes, absolutely. As I point out in the video, with the F6, you’re limited by the analog input circuitry. But you’re no longer limited by the recording medium, and recorders with more dynamic input circuitry could be built in the future.
@@simonpeck I got that part, but how much range do you need above a set limit (mic)?
I definitely agree on recording a higher resolution and headroom needed for the endgoal, always. But if if the air itself, without the mic, only can go to 190db? Hmmm this reminds me of using an amp with 10x the power you actually need for the speakers. Those speakers are singing like never before. There is some merit in what you say, intuitively. Fascinating.
Whoa never knew ! Cool!
I’m not sure I understand your point. Clip gain has been a thing for a very long time. 32 bit float, as I understand it, allows you to clip internally in a DAW without actually clipping the output.
This is so much more than setting clip gain. It's about capturing the full dynamic range at the time of recording.
Are you saying that
24 bit is too risky to record Drums?
Nope! This is another technique that one could use to record drums that has certain advantages.
@@simonpeck could you outline the advantages please?
I recently got one of the zoom live track recorders and they 'only 'go up to 24 bits
Probably wrong: with 32 bits you are able to represent 2^32 distinct values no matter if integer or float representation is used. Float allows wider ranges, but you lose resolution.
Yes, sorry I explained that incorrectly.
Zoom released a 32 float audio interface, use that instead. Attach 1 or 2 mics of your choice to it and record it with whatever device, laptop for example
I suppose I could get 6 of them and create an aggregate device in Mac OS to support all 12 microphones on the drum kit.
@@simonpeck if you attach a 5v powerbank to each of them then they have enough power to run without needing to attach them to a power outlet in the wall
Modern way of drumrecording is using the Yamaha EAD 10. If you want a quick work flow ...that is used by many modern drummers. Don Sanderson records his drums 2023 using the EAD10 . www.youtube.com/@DonSandersonDrums
I was really enjoying this video then I realised you’re a small channel, really punching above your weight! Keep it up!
I’d love to know more about your whole setup, I just bought my first proper mic (sm57) to do some recording for a project after only using DI bass and guitar and I’m completely hooked. I’ve done some bits at college with their studio but Im looking to get a decent setup going at home.
Thank you!
Anyone know what side address mic that is at 0:08?
That's the sE Electronics VR2 ribbon mic. I'm working on an upcoming video comparing 4 different ribbon mics at wildly different price points.🙂
I actually tried something similar with my Apollo (24bit) and didn’t add any ‘gain’ to it (i think it defaults to 10db of gain?). Made for an interesting experiment. After recording i ran it through actual preamps for fun. Digital and analog. So you can hear the ‘sound’ of the preamp, but something irked me a bit with that test. I think this recorder idea would make it even better. I also have been seeking something to do more remotely and i may just use this if it doesn’t require setting levels.
Yeah, it's really freeing to be able to walk in and hit record. One thing I didn't mention in the video is that the recorder can do 24-bit and 32-bit float at the same time. So you can set your levels the way you normally would for the 24-bit recording and decide which tracks you want to use in post. Also the Sound Devices MixPre 10 II can act as a 32-bit float audio interface if you wanted to record into a DAW.
Subscribed! Great stuff, Simon. I‘m recording my drums with a Presonus Studiolive Mixer as an Interface into Studio One. I have to check out if recording 32bit float is possible within this setup.
"I exceeded the input capacity of my audio interface so I used a pad on the xlr cable."
"Levels? Wdm levels I record in 32 bit-float. I don't use "gain" or things like that"
The weak link in this setup is definetly the preamps, but it's less about the gear and more about how you use it!
You can lead a horse to water…
Very interesting! I started learning drums with the same idea. After 10 years I am still not a really good drummer. The more you learn the more you realize that drum playing is almost an infinite task. as a sideway I learned how to program drums. With this I‘m quiet ok. When I record my live drums I mice them with 4 to 10 mics depending on the sound I want to achieve. I have a formula how I process the drums that works for me. Just for the sake of learning: I would be interested how you record the drums are and how you process them.
Thanks, Michael! I'm working on a video to illustrate that.
I love working at 32 bit... a lot more flexibility after the fact.
Very interesting. I use Cubase12 which I can record at 96Khz./ 32 bit floating point. I do this when I record projects that need to have beautiful detail in tone, reverbs and FX in general. Generally not loud rock tracks, as the detail isn't useful. For them I use 48Khz / 24 bit. I leave this comment only as a reference. I hasten to add: The song is the thing. Good songs well played tend to make good recordings. All the Best.
Whatever interface you are using it can pretty much be guaranteed that the A/D converters are not actually 32 bit float. The converters are probably 24 bit but the data can be stored as 32 bit values. Whatever format the values are stored doesn't actually translate into a better resolution than the number is being represented by the bit-width anyway. Hence why the industry still uses bit width to define the resolution of recording and not say use the internal storage representation. For example, the numbers could easily be stored in 64 bit float but that wouldn't increase the resolution the audio was recorded in. Once the data is in the computer it makes sense to have the highest practicable floating point representation hence internally DAWs (like Reaper) use 64 bit floating point.
Note, even at 16 bit resolution you probably have more than enough dynamic range to record things like drums even allowing for decent headroom to avoid clipping, -18 dBFS is recommended usually, which would give you 80dB of dynamic range!
Zoom and Sound Devices implemented it by using dual 24-bit converters on each preamp. The problem this solves is that not only do you not need to set levels, you can't set levels. It isn't about having enough dynamic range. It's about eliminating a step from the recording process. You can set the level in mixing, after the fact.
@@simonpeck I see. It's not obvious from the specs of the device but if it has dual 24 bit A/D's it might mean it has 48 bits of resolution. This would mean you actually have a massive dynamic range. The would then just set the mic preamps at the lowest levels to gives some level which is above the noise floor.
It uses one set to high gain and one to low gain and has to make an intelligent decision about which to use and how to combine them where they overlap.
I’m interested to know how your workflow works with this? For example if you wanted to play with a click or scratch track how do you do this? I’ve been interested in a concept like this since I got hip to the sound devices line of products as I’m in the same situation with my home studio being somewhere that I can’t track drums in. Thanks for this stoked to see more!
Sure, I'll try to illustrate that in an upcoming video.
So how do you sync the drum takes to the music?
What about multiple takes of the same song?
Audient do an interface that does auto level for drums. You just play a bit and it figures it out.
Playback and/or a click track can be sent into one or more of the channels in one of the recorders. Or it can be done manually. Multiple takes have to be done serially with this method, so that is a potential drawback. They can then be handled in the DAW.
@@simonpeck yeah ok. Seems like a lot of extra work to me. If I was recording drums remotely I would be getting a 16ch EVO and a laptop and using my phone as a remote.
I do love finding great non studio drum spaces to record though. Some of the most legendary drum sounds have it been recorded in studios.
The Sound Devices MixPre 10 II can also be used as a 32-bit float audio interface if you wanted to record into a DAW. And yes, this is a great method to record in unusual spaces.
REAPER records at 64 bit float by default. I know from skool that 64 is more than 32. This is terrifying. Is my machine gonna blow up? Or maybe my tracks will mix themselves? Help
Well Dang!! Love it!
What do you mean run the tracks through "preamps later"? You can run 16bit audio through outboard gear later, you dont need 32bit float for that. Also the normalizing can be done the same. The noise floor of mics and preamps themselves limit the dynamic range, and all this 32bit float saves you is the trouble of setting a gain, right? Maybe explain how it's different than a decent 24bit recording setup, besides saving 1 minute setting gain. Have you recorded traditionally before?
It saves a lot more than a minute when you’re using 12 mics, especially when you’re the musician and the engineer. It removes an opportunity for human error.
Yes, I record vocals and other instruments traditionally in my home studio.
I didn’t mean to imply you can’t run any audio through outboard gear later. The difference is that if you’re in a situation where playing drums is prohibitively loud to play in your home studio, you can record them remotely in a simple fashion (without even using a computer) and handle the rest of the workflow in the home studio.
@@simonpeck thanks for your reply, I guess I'm still learning about this so it helps. I do get it from a workflow stand-point, was really more curious about if there's technically an advantage in terms of dynamic range/resolution. No need to explain further, I've since read up more about it. Rock on...
🤔 32bit float, eh? Well, I've been in that wagon since Cubase VST32 1998 or 99. ✌️😎
The original 32bit float DAW since last century. Never had to worry about my mix clipping and also provided none destructive editing processing. Everything you do in Cubase or Nuendo is on a 32-bit float audio engine.
Now everyone is trying to catch up on this.
Most DAWs use 32 bit float internally. But that is different than capturing the audio this way. If you record through a 24-bit audio interface, you’re limited by that no matter what format you’re working with in the DAW.
Hate to say it but.. In this context, 32 bit, float or not, is just a number used for marketing. Due to the noise floor limitations of the preamp (as you touched on) you’re not gaining any more usable dynamic range than a regular 24 bit ADC. And trust me, 24 bit is already a TON of dynamic range.
32 bit is more useful for post processing “in the box”, or in digital mixers. This allows extra headroom to prevent clipping from oopsies in the mixing process, AFTER the signal is digitized at 24 bit.
I would suspect you could use the same workflow as your zoom setup, with a 24 bit system, and achieve the same results. Which is fine if it works for you! But 32 bit isn’t achieving anything fancy.
For a portable setup, I would usually use a portable digital mixer as my interface (Midas MR18, or a soundcraft UI24) These function as 16+ch interfaces to connect to a laptop, and allows easy remote preamp control via an app. This way you’ll be operating in the sweet spot or your preamp, and getting maximum actual dynamic range.
The marketing is confusing, because it’s the dual preamps and converters that differentiate the recorder more than the data format. Sure, 24-bit has enough dynamic range, but this method eliminates gain controls altogether.
@@simonpeck Oh, so 32 bit is just used as a method of containing the dynamic range of dual-preamps. That makes sense- then you can get 24 bit quality after digital gain later. That’s actually kinda neat.
32 bit is definitely not the hero then, although is albeit more space efficient than recording two separate 24 bit files, one for each pre.
Regardless, I enjoyed watching through the entire video, and understanding how floating point works. Thanks!
This is so strange, why is his voice so monotone?
Mmm….kinda.
You don’t have to worry about levels WITHIN REASON. Not just willy-nilly.
Prince set an incredibly high bar for hit songs, and the drums were all lowly 8-bit samples from his Linn LM-1.
Do really good drums mic’d with great mics 👉potentially👈 sound really interesting? Sure, maybe. In some instances.
It shouldn’t be overlooked how much your smart decisions with what you know about drums and mics play a role. Do this with lower bit recording and it will still sound 95% just as good. At least.
I can do a different smart thing - I can take cheap drums and put brand new inexpensive Ambassador drum heads on them, and then use nothing but cheap Shure SM-58 mics and I get killer recordings even just at 44.1khz 16-bit, (CD Quality) because I know how to best tune the drums and where best to place the mics.
Note: Put distorted electric guitars over ANY of this and it’s all OBLITERATED.
That’s precisely why we love distorted electric guitars!! 🤘😆🤘
Using 32-bit float just means we have more headroom.
So we can worry LESS about the noise floor creeping into our sound, and we can worry LESS about hitting something too loud and causing digital overload breakup.
If the sound you create is crappy sounding, all it’s going to do is give you a super clear recording of a crappy sound.
Clearly a lot of research and learning was done here in order to come up with this very clever and very effective recording process.
I’m just arguing that one can use this type of a setup whilst recording at a lower bit rate, and the results will be nearly indistinguishable for most listeners.
I personally do not find 32-bit float to make much of a difference in the recording process.
But in the MIX process it’s a game changer.
The DAW called SONOR(now called Cakewalk and given away FREE) was using 32-bit floating point processing for its internal processing and mixing YEARS before other DAWs were, and this was what gave it such a sonic advantage.
Is 32-bit float better for everything? Mmm, not necessarily. But it’s never worse.
Is 32-bit float just better? YES.
These recorders use dual fixed-gain pre-amp/converter stages (high and low), so you can't adjust the levels if you wanted to.
Of course you need to set the levels and sound, how else are you going to make sure the mics are in the correct place? Get it right at the source rules apply in all forms of recording. 32bit float makes no difference to the process.
Recording in 32 bit float changes the process because (not only do you not need to set levels) you can't set levels. The entire dynamic range within the limitations of the analog input circuitry is captured. In terms of mic placement, it's still possible to monitor the output. Nothing about this process goes against the philosophy of getting it right at the source.
Se você equalizar e comprimir um pouco durante a gravação o som vai ficar muito melhor para processar depois.
👍👍🏻👍🏼👍🏽👍🏾👍🏿
Interesting stuff! You should learn to tune your drums though if you want pro recordings.
What was out of tune to your ears? The floor tom? I’m pretty obsessive about tuning but I think it was a little off that day.
Pretty cool info, is this channel working with AI to produce these?
No! What makes you ask? Do I act like a robot? 😂 I did use Midjourney to assist with a few thumbnails, but that’s the extent of it.
How is this honestly different than using an interface, preamps, etc in a “traditional” setup, and recording at 32float? You’re just replacing the interface and preamps with your zoom. Maybe I’m missing something.
The main differences are:
1) It saves time -- There's no need to set levels.
2) It reduces the risk of human error -- There's no need to set levels.
3) Drums can be recorded in a remote location without even bringing a computer. The home studio can remain at home.
4) More of the recording budget can be spent on drums and microphones where the impact will be higher than on gear further down the signal chain.