I know how to set levels, but this is objectively better. Audio is unpredictable. People make mistakes. This solves both problems. If you're a musician, just have this thing recording every time you sit down to practice or jam. You might capture some magic and the levels will be fine no matter what.
Exactly, it's a HUGE quality of life improvement. I also noticed that using this changes how I use a lot of my microphones. It's no longer about which mics capture the most information within that narrow decibel band during recording. It's about the tone quality that I get _after_ I adjust the levels and compress it. Some passive dynamic mics are able to give me much more detail and fuller sound than my condenser mics in post. Also, no longer needing _extra_ preamps on things like ribbon mics is a huge win in the battle against the noise floor. Biggest drawback, though... you gotta be really on top of your room treatment. You're also gonna become keenly aware of the inherent noise in your setup.
that is pretty badass! as a guitarist, there are so many smaller interfaces that just cannot handle really high output pickups, even if you set the input gain to minimum. so you'd need to mess with DI boxes or find an interface with a pad switch. this seems like a really good interface overall, especially at the price. 5ms-ish RTL is acceptable to me.
You have an amazing channel Julian. As an engineer in electronics, it is refreshing to find a technical channel like yours, among many RU-vid channels of "producers" who it is pretty obvious that do not have a proper degree to talk about technical subjects such as digital analog conversion, component specifications and performance measurements. It would be great if you could get an Apogee Duet 3 in your hands to see your interesting analysis. Cheers
Great video as usual. Of course I pulled up the manual while I was watching, and the inputs can take +24dBu. Although that's enough headroom for almost everybody, I will attest to once getting a signal that hot, when I streamed a music festival on the island of Sardinia in 2019.
On MacOS Core audio supports 32 bit float and even Apple's own ARM based laptops have a 32 float converter for the internal microphone for Macbooks. Even though MacOS supports 32bit float recording though out the OS but there are certain applications do not. Logic Pro and Mainstage does not yet support recording or inputing 32 bit floating point audio only 20 and 24 bits. But Reaper, and Pro Tools do support it already. As well as MacOS's screen capture software and Quciktime recording does support it. I think this years OS update will have full 32 bit compatibility across all Apple made Mac Apps that support audio recording. The Logic/Garageband/Mainstage team have said 32bit float recording will come before the end of the year to those app and it will probably come to final cut very soon after that! Cheers Julian always great work!!
I have a Zoom M4 that records in 32 bit float and has a USB interface mode with 48K/32-Bit Float. However, in Reaper on Windows, it is an unusable nightmare currently. I'd like to blame Windows...but I have a feeling it is simply the firmware itself within the M4. I'll try with my M1 Mac and edit this comment soon with the results.... EDIT: Seemed to work just fine on my M1, make sure to go into Project Settings and under "Media" make sure WAV is set to 32 Bit FP. Still no luck on my Windows Laptop. Could just be my very new laptop is waiting for more updates (2023 Gigabyte Aorus 17H) EDIT2: Zoom M4 48K/32-Bit Float works flawlessly in my Windows 10, Ryzen 3900X, RTX 3090 desktop PC in Reaper. My laptop must be so new that it needs more updates.
@@GeekTherapyRadio It might be the CPU if its AMD, maybe look to see if there is a Firmware update for your CPU. I know AMD Ryzen is notorious for audio issues and its such a shame since they are super good cpus. I know my Antelope audio interfaces have 32 fixed point converters that work at 32 bits on MacOS but on my Windows and Linux laptops it only will show 24 bits unless I use a usb 3.1 port or higher so maybe try your other USB ports.
@Music Factory Studios Awesome suggestions. Laptop in question is a 13700H Core i7. Latest. It has trouble with every interface I've used. I'm chalking it up to the chipset needing updates. My AMD 3900X CPU works flawlessly with my interfaces, including 32 Bit Float from my Zoom M4. But it's a mature platform. Hopefully Gigabyte will role out chipset updates for the Aorus 17H. It's still a bit buggy.
As mentioned in the video oc it can still clip... people should not forget that 32 bit float is just a "storage format". Everything before that is still hardware and has a physical limit. Starting from the microphone that has a maximum peak pressure all the way down to the ADC that has a maximum voltage input
The design for FS= +24dBU at the input connector makes that, if not impossible, at least practically so. Power supply voltage is the limiting factor. Sources would need a 40+VDC power supply, almost unheard of today. The key is that any conceivable operating scenario will be physically unable to reach that voltage, hence no input gains.
@@artysanmobile It seems like you are putting artificial limits on what you think is possible. Having designed and built amplifiers I feel like I could overload any input easily. Also, the PSU voltage doesn't have much to do with what the output voltage is capable of. The circuit can always multiply voltage above whatever the PSU voltage is. And why do you think 40V DC is some kind of limit? What if I hooked my vacuum tube amplifier up to the input of this audio interface? It has a PSU voltage of over 800V DC! lol
Yes obviously, but in practice, it's almost impossible to ever reach those. For example, the average maximum peak pressure of a condenser is around 130dBs SPL, which is the sound of a shotgun at 3 feet (yeah I know, this isn't a very scientific thing but that's the best we got lol).
@@mjodr ...No one is saying that it's impossible period. We are saying : unless you specifically wanted to, used a variAC to drive whatever you're gonna input above limit, or made whatever you're gonna input clip by default otherwise, it's not gonna happen unless you want it too. I mean, statistically, the possibility to clip if you don't want it tom is so low, that it's virtually impossible. No one is saying that this is impossible in the absolute. If you want to set the thing on fire... you can
Hey Julian, I would love to see you do a video on the Zoom F3. While I know it is similar to the UAC-232 and likely uses the same ADC schematic, it includes for example, a 24-bit MME driver mode for plug and play interface capability, which you asked them for in this video. Thanks as always!
Terrific video, Julian! So 32-bit float audio with two ADC's is somewhat analogous to High Dynamic Range technology in cameras, where two images are taken at two different sensitivities in order to capture detail in both bright clouds and dark shadows. The end goal with both technologies is to capture the dynamic range of the real world without clipping and then eventually compress the information into smaller range for playback: 1.6 billion nits for the sun compressed down to 2,000 nits of brightness on your iPhone screen; or 190 db of sound compressed down to a CD-quality stream for your earbuds.
FL Studio can record 32 bit float into Edison and into the Playlist. I believe this is ASIO or CoreAudio only as these are low latency drivers, MME (used to be 16bit until Windows 2000 or something ) and other internal cheap stock soundcard solutions will typically be 24/32bit either depending on the source and can be weird sometimes not allow the passing of data samples as processed from the bus (USB port) but instead as CAPTURED (via the driver). Make sure your drivers are 32bit float and updated, as well as your recording software. p.s. apple users, I think Logic is 16bit until you hit 24bit BWF recording etc, I have no idea if 32 bit is a thing or is planned. I rarely use Logic other than for mixing.
Most excellent review! Been waiting for it and you did not disappoint! Next up, can you please review the SoundDevices MixPre 3 II? Which also has ASIO drivers. Thanks!! You're the best!
I think the biggest thing for this interface is super hot active guitar pickups or active bass preamps. No internal clip and just setting the input properly in the app, I'll be good to go. The Focusrite 2I2 second gen can't handle hot pups without mild to moderate cliping. I generally don't care as I play metal, high gain ampsims, but it does effect cleans and edge of breakup tones. Excellent video.
I have my eyes set on the 8 channel rack-mountable version Zoom will release later this year! (UAC-832). Looks like Zoom are taking notes on UAC-232's users feedback to improve the tech and features. They should hire you for their Consulting Team 😂
@@avocadochoe They have a Clock Source selector (with "ADAT" being one on the options) on the front panel picture/3D mockup on their website, so I'm guessing yes!! I think it's still in development, cause I haven't found anything anywhere else.
@@sonidojamon Oh my I am very excited for this! Hope they flush out some of the minor down sides mentioned in the video before the launch and I'd be sold! Should make an excellent home recording environment.
Thanks for the amazing review as always Julian! Hope to see you eventually get your hands on the new Boss Gigcaster digital mixer / interface to see if they are any good!
The UAC-232 owner's manual actually explicitly states that they use two DACs, one low sensitivity and one high sensitivity and blends the two on pages seven and eight. zoomcorp.com/media/documents/E_UAC-232_.pdf
For hi-z direct guitar this is important. Early entry level interfaces drastically lacked the dynamic headroom for things like active pickups. I've recently been using a Vox ValveEnergy "Nu-tube" pedal as a (clean-ish) preamp into digital amp sims. IMO it "feels" more authentic and makes using analog pedals direct noticeably more convincing.
"Huge dynamic range" is a bit misleading; the maximum effective dynamic range of a 32-bit float is exactly the same as a 24-bit integer (due to the sign + mantissa being equivalent to a 24-bit integer); it's just that the clipping point (i.e. the maximum integer value before overflow occurs, usually represented by 0dBFS) is able to "float" to higher or lower values as the exponent changes to accomodate louder or quieter signals. If all you wanted was the largest effective dynamic range, a 32-bit integer would beat a 32-bit float; what's actually happening is that in exchange for the extra 8 bits, we get protection from clipping at 0dBFS rather than an increased dynamic range.
You missed a detail in the float format. While it stores 23 fraction bits in the mantissa, there's an implicit 1 in front of them. The integer precision range matches a 25 bit integer when you're including the sign bit in both. The format functions like switching full scale range per sample to the lowest linear doubling which does not clip and recording that gain level in the exponent. Of course, just because the format supports 25 bits of precision doesn't mean your ADC can; with two overlapping ADCs, your best range would be just below clipping as usual, but there are two places for that. The higher amplitude ADC offers a lower precision fallback, so you'd target the low amplitude ADC's peak level. Having that fallback could let dynamic gain be useful (misses wouldn't be disastrous), if the gain levels are precisely calibrated to the large scale range.
I purchased Rode NT1 5th generation which has 32bit recording capabilities. I still do not understand how it works other than it has 5 AD stages so it has a lot of dynamic range, but how it stores that data is beyond me at the moment. :) In the same time I really think this is a gimmick for many users, not sure who would need this except those who need to record something on the fly when there is no time to set the input level, but in the studio this seems like an overkill.
Lightning strikes in the studio, not just during field recording. The talent unexpectedly kicks on a pedal part way through and creates a great moment, but you can't use the track because it was too hot, and they clipped your interface... 32-bit float seems like it could be a timesaver (lifesaver?) in that situation, but only if it truly works as advertised, and your DAW co-operates.
@@steverolfeca Of course you could think of many such situations, however this is really a minor concern. I know many people having problems with setting the input gain properly, and the RU-vid is full of instructional videos showing how to do it, but in my reality this is a 3 seconds task and I never had any problem with this, nor I understand what is that fuss all about. It works well, I tested it in Reaper and sure the dynamic range is there. Maybe you are right, maybe this is the future. I remember many complained about Reaper not having separate midi, audio, group, and aux tracks, but only one type of track to be used for everything you need. It is a brilliant feature, maybe 32 bit recording will feel the same soon, I do not know really, but for now it looks a little bit gimmicky to me personally.
@@steverolfeca You have to understand that even though it happens the same time in the product, the trick of merging two different adc stages has nothing to do with 32bit float. Even with the trickery the dynamic range is still only 130db or so which is easily contained in 24bit fixed point. And the current 32bit fixed point availible in many audio interfaces has 192db dynamic range , this would be more than future proof for audio recording.
You find tutorials on YT about floating point. I wouldn't be able to explain it. For audio purpose, no clipping because you can exceed 0dB fs and accuracy tends to be infinitely good, you can exceed 10^-38 up to 10^-45 values, very small values. Though adc should be integer, that's a lot of wasted space
when i record radio ads at the radio station the 'lawo ruby' listens for the first ten seconds and sets the perfect volume for my voice or anyones voice.
Good grief, I'm still using Cool Edit Pro 2.1 from around the early 2000's, and I do everything in it using 32bit floating point format. Of course, CEP was way ahead of its time, and IMHO much better than what Adobe replaced it with (Adobe Audition), but it wasn't really set up to support the more modern workflows that have evolved since, especially when they rely on GPU's. And no other ADC has yet output 32bit floating point for recording, but I would record from any normal 24bit ADC as 32bit float in CEP (and of course, you have to set the levels to avoid clipping, but with 24bits that isn't hard to do). But I'm astonished to hear that people are still using software that cannot record in 32bit floating point format! I mean, really, I've was doing it 20 years ago with CEP!
The more I learn from you about 32 bit float and merging two DACs, the more skeptical I become, wether this would be of any use to be. My current setup consists of a MOTU M4 Interface and a Shure Beta 58A dynamic microphone. I have set it up for about 25dB of headroom over normal talking. And my SNR at 64dB even after 21dB additional gain via a maximizer plugion vs Selfnoise is already extremely low as is. So... I don't know... Dual DAC and 32bit float seems to be quite gimmicky, at least for most uses. Stil.... adding that much dynamic range is quite a nice job.
@@GuyGamer1 I’m from back in the day when they used double system location recording in movie making, recording on a nagra (reel to reel) with some sort of mixer between the mics and the nagra. The crew position was called “mixer”. The mixer was responsible for getting the best audio recording possible, which as you can imagine was/is pretty important. Good mixers were paid well. Bad audio can be fixed in post, but back then post production processing was expensive. Mixers (and their boom operator assistants) were not to be trifled with, as they could listen in to private conversations. But they were also in the know with all sorts of useful insider information, such as when’s lunch? I don’t think we regular crew people gave them enough credit or patience. We’d grumble about delays while the mixer chased down ambient noise, like a refrigerator in the next room. We’d groan when they had to track down the source of line noise. “Hold for sound” is not what you want to hear when you’re one or two takes away from lunch. Anyway, these days the demand for people that “know what they’re doing” far exceeds the number of people with those skills, no matter the level of production. While 32-bit float is going to be perceived as a magic bullet for many “content creators” (especially solo creators and small teams), the need for skilled technicians such as yourself is not going away. The float is just another tool.
Single handedly saved the music hardware part of the feed from another MPC X LE video lol Question/Concern: "Everyone's" argument usually goes something like "32 bit is unnecessary, cant reasonably capture or perceive that range, the file size!" & some other Hot takes... Answer: "Cool, don't use it then... 😆" Seems to be good for non-Audio Engineers, especially those with a noisy home studio or those who record out and about with mobile setups (outdoors or even all types on non treated rooms ie hotels & so on) Of course the mic/outboard preamp situation is a big factor but simply capturing everything after that fact enables ONE to do the rest in the box without jumping through a labyrinth of hoops Another interesting product is the Rode NT1-A 5th Gen USB/XLR mic. The catch is that the 32-bit float recording is built into the mic and is only possible when using the USB port. XLR is only 24 bit. Regardless would be interesting to see a side by side with that mic on USB (32-bit float) and then running through the 232 via XLR at 24 bit 🤔
If any of you guys have bought it, please review the headphone amp sound characteristic compared to any of your older audio interface and other devices.
Can you review the new Newman MT 48 interface? It has a very similar dynamic range, so it would be very interesting to see the similarities and differences given the huge dynamic range of the UAC-232.
Thank you for making a good video. I'm thinking about using UAC-232 as a portable audio interface. If bus power is possible, it is attractive that 32-bit recording is possible just by connecting to a MacBook. (Although F3 alone can solve the problem, I think it can be used as an audio interface that supports balance output.)
Great video as always!!, a year ago you made a video about how to plug your headphones to your interface so I followed the instructions and it works perfect but you never showed how to unplug.
It's weird hearing "is 32 bit float the future" like it's something new. 25 years ago 32 bit float was available in Cool Edit Pro, and eventually Adobe Audition who bought CEP.
Well nice, but I am not 100% sure if the switching is not audible in complex musical tasks. With measuring tones, you can not get the same results as experienced pros will hear in practice. And I am not convinced that this sounds the same as a regular non switching amp/ADC combo.
Thank you for the review! I appreciate the part about 32 bit float compatibility with different capture programs such as OBS. Do you think you could try the obs-asio plugin with OBS to see if that fixes the issue?
...No one is saying that it's impossible period. We are saying : unless you specifically wanted to, used a variAC to drive whatever you're gonna input above limit, or made whatever you're gonna input clip by default otherwise, it's not gonna happen unless you want it too. I mean, statistically, the possibility to clip if you don't want it tom is so low, that it's virtually impossible. No one is saying that this is impossible in the absolute. If you want to set the thing on fire... you can
Also 32b-fl requires your DAW (or any audio software) be capable of handling 32bit data, which apparently OBS & Discord (& likely many other apps) are not, so they still clip.
@Julian Krause, great review! I'm so disappointed to see that ZOOM made a new driver that has an average performance. I have a Zoom UAC-2, with is among the best USB interfaces performers! The RTL latency is at pair with RME but this one seems to have a different driver/engine. See Below: Sampling Rate: 44.1khz ----------------------- Buffer Size: 64 samples Total RTL: 5.49ms Buffer Size: 32 samples Total RTL: 4.04ms Buffer Size: 24 samples Total RTL: 3.67ms Sampling Rate: 48khz ------------------------- Buffer Size: 64 samples Total RTL: 5.04ms Buffer Size: 32 samples Total RTL: 3.71ms Buffer Size: 24 samples Total RTL: 3.38ms Sampling Rate: 96khz -------------------------- Buffer Size: 128 samples Total RTL: 4.47ms Buffer Size: 96 samples Total RTL: 3.80ms Buffer Size: 64 samples Total RTL: 3.14ms Buffer Size: 32 samples Total RTL: 2.47ms Observations: 1) I tested the performance by playing Reason Demo Song. I'd say it is "low/medium" load. I also tested it using a stand-alone version of Amplitube. 2) Using an OLD Core 2 Duo @ 2.4GHZ computer, with Windows 8 and connected to USB 2.0, I only got some clicks using the 48k@32 Samples and 96K@32 samples 3) Using my main laptop, I7-3630m, 8GB RAM, connected to USB 3.0, I got NO CLICKS or POPS WHATSOEVER. Yes, the computer is optimized for Audio and it is pretty clean. 4) Using the SAME i7 Laptop, but connected to USB 2.0, I DID GET SOME CLICKS in the lowest settings, but less than with the "old" C2D laptop. 5) As reported, there is just a VERY slight RTL latency difference when connected to USB 2.0 vs USB 3.0. Don't remember now, but less than 0.2ms., so I did not even write it down..... Conclusion: I had so many interfaces during the past few decades...PCI, FireWire, USB....the ZOOM UAC-2 is BY FAR the MOST STABLE and lowest RTL latency I even tried/saw in person. Yes, I never compared it to an RME, so I leave that window open...but as a Portable/gigging device, it is a no-brainer. And I hope the other manufacturers learn the lesson because it is incredible that in 2018 we are still suffering from this...nothing worse than an unpredictable interface...no one can play live and feel relaxed if there is the slightest change of interface crash. gearspace.com/board/music-computers/618474-audio-interface-low-latency-performance-data-base-70.html
Absolutely agree with you here on the need for better software. Far too many manufacturers go on about specs and give scant regard to the software. Never tried the UAC-2 but the software performance always seemed remarkable on paper and plenty of user experience reports out there to say those figures stood up in reality. Reliability is why I went with RME (Digiface and ADI-2 FS combo).
I use a Zoom F6 to record piano recordings and it definitely can clip and distort. Piano is difficult to record. The secret is to set levels lower than you think. This is great technology, however the “never clip” is being wildly oversold. No such technology exists.
32-Bit Float is miraculous for ribbon mics (or any low output source...or high output for that matter). No need for inline gain boosters that potentially add noise. I just recorded my cheap MXL R144 with the Zoom M4 for this video, if anyone is curious. No Cloudlifter. Just straight into the M4 ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-q-WC25no-Oc.html
Hi, what do you think about comparison of Zoom 232 and Arturia MiniFuse? I have an Arturia MiniFuse and looking for more urban variant. I think that 32 bit float technology is very useful, and I hope that all manufacturers will be doing such interfaces in the future.
Thanks so much for your great reviews. I wonder if you have heard about iConnectivity AUDIO4c due to its capability of being shared simultaneously by 2 different devices. Would love to see a technical review from you. Cheers! ✌
How can it be compared with zoom f6? zoom f6 has more versitile audio interface or field recorder but there is no audio latency information..In m1 macbook air and m1 mac mini logic pro the least i/o buffer size ,32 the RTL is 7.8 ...
Julian, I would like to see a review of Roland interfaces.The old Quad and Octa capture interfaces were highly regarded because they had a very good mix of hardware and software. I would like to know if it's still the case.
Finished watching the video. It's great that you mentioned it's still possible to clip and the dynamic range is still within the 24bit 144db. But I think in the verdict section one thing should be said clear(imho) the trickery that combines two adc signals with different gain has nothing to do with 32bit float. The dynamic range is still within the capabilities of 24bit and would work more than perfect in 32bit fixed point format which has 192db dynamic range so this completely covers future possibility of actual 130db adc plus 30db or so analog gain(160db is probably the highest actual meaningful and achievable dynamic range needed).
Hey Julian, can you test the zoom f3, tascam x6 and x8 Porta capture and the steinburg ur44c? These are 32bit audio interfaces. But I am unsure if they have low pre amp noise or if the portable recorders can actually be used on a computer and properly capture 32bit float. Thank you so much for your videos! I wouldn't be able to pick out a good audio interface without you! Edit: so the steinberg ur44c is 32bit integer not float. I really want a 32bit float interface with 2x independent headphone monitoring Jack's.
Zoom UAC-232, Steinberg IXO22, and Steinberg UR22C are close in price (not far off 2万円) where I live. I am just a RU-vidr using XLR mics for video voice over and podcasts. I don't yet know how to use a DAW. I currently have an Evo4, but the USB C port disconnects if I touch the USB C cord. I then have to jiggle the cord, mess with settings, and restart Davinci Resolve. I am looking for something new....or something to use as backup. Any ideas anybody?
Nice video. I think 32Bit float will definitely will be the future of audio recording even though we are not quiete there yet. Besides that, I would have a suggestion for a future video. The Rodecaster Pro 2 interface/mixer claims to be the perfect solution for livestreaming and having very good preamps. Addionally it has a lot of options which could replace for example 2 interfaces, a DBX286s and a mixer. I could not find any measrements on the internet, so maybe you would like to review this.
Some people are having issues with the Rodecaster Pro 2. Some of them have an issue with the plastic separating from what is under it, some of the square buttons have failed and Rode doesn't offer much info about those issues, although they may replace them under warranty. But what happens where they are out of warranty?
24bit format of 144db dynamic range is more than enough for future analog circuit and adc conversion. 32bit non float pushes this to 192db which can capture noise down to non physically possible level. The float thing is completely useless.
@@johnyang799 Except real-world 24-bit converters only really have ~18-20 working bits max. 32-bit converters are functionally impossible. Maybe stacked ADC's like in the Zoom will be a good workaround. Time will tell, but right now I'd stick with a traditional ADC setup.
@@erinburke9711 No no. It's true that a currently good adc is 20bit-21bit or so dynamic range. In conjunction with the same dual trickery and combining the two signals you'd achieve the same high 130db+ dynamic range. However this isn't issue free especially there are only two levels basically. Which is the same as there is only a gain toggle between 30db and 0db instead of a gain knob which is good for majority of the range. But it's not noise free nor has good SNR which is usually indicated by dynamic range(not here). This technique is still good for the most part tho. Normally you only gain 3db of SNR and DNR with 2 ADCs but here you can get 25db or so extra DNR. Rode has 4 ADCs so this is even better. All above is not required on 32bit float. If you combine the good adca with gains you can get 130db + 30db dynamic range, even then it's contained in 32bit fixed point. And this results in no 32bit float compatibility issues.
If they put multiple headphones outs or had 4 Inputs like the AMS 44 i would brought it for live streaming my music performance from home.....i hate the trade offs here and there that could've made it a great solution